Re: XAUI AC coupling
With reference to the question: "2. What is the impact of this decision on
low jitter design?" which Jonathan Thatcher asked in this thread. Here is an
example of where AC coupling can improve timing margin of a signal compared
with DC coupling.
When I mension "duty-cycle distorted data" I mean the zeros are wider than
ideal and the
ones are thinner than ideal. Or visa-versa. This can arise from the last
stage of multiplexing in the transmitter particularly if the transmitter
uses a half-rate clock which does not have a 50/50 mark/space ratio.
Consider a XAUI data signal which is duty-cycle distorted such that it
spends 60% of the time at the zeros level. The signal is subject to slew
rate limiting by the transmitter output circuit. Channel bandwidth
limitations furthur limit the maximum frequency content of the signal. The
signal arrives at the receiver with finite edge speed and marginally meets
the timing specification of the received eye. If AC coupling is used, the DC
content of the signal is removed on the receiver side of the AC coupling
capacitor and, due to the finite edge speed of the signal, the duty-cycle
improves as a result. Approximately half of the duty-cycle distortion is
removed and hence timing margin
is improved. The disadvantage is that voltage margin for the zeros reduces.
Hence in this case, AC coupling helps with timing margin.
-Paul Wilson
-----Original Message-----
From: Rich Taborek <rtaborek@xxxxxxxxxxxxx>
To: HSSG <stds-802-3-hssg@xxxxxxxx>
Date: Saturday, September 23, 2000 7:43 AM
Subject: Re: XAUI AC coupling
Jonathan,
Please allow me to simplify this discussion. If I offer a customer an
XGMII extender consisting of two XGMII-to-XAUI chips, I would likely
offer the customer two of my chips rather than one of mine and the other
from a third party. The two chips are likely to support DC-coupling by
the mere fact that they're the same chip. This is clear proof of XAUI
DC-coupling. I don't know where you're headed with your argument of an
AC-coupling requirement. I'm merely stating that XAUI, an XGMII
extender, should allow DC-coupling. Therefore, the converse is true: The
P802.3ae standard should not require AC-coupling. It's as simple as
that.
Let's quit beating around the bush. I've already put up the challenge.
If it can be proven that AC-coupling is required for XAUI
interoperability, I'll back down and change my tune. I've discussed this
issue within nSerial and with outside experts before posting my initial
note. I achieved 100% agreement to not requiring AC-coupling for XAUI in
the standard.
The only question that should remain is whether both AC and DC coupling
should be specified in clause 47. I've addressed this issue in a prior
note on this thread in response to Dawson Kesling.
I'll try responding to some of your other questions below. Am I
qualified? ...only as a novice in this area. How does a novice become an
expert? ...by sticking their neck out and getting it hacked at by the
experts :-O
Jonathan Thatcher wrote:
> Rich,
>
> If mine is twisted, yours is a full double helix :-) Let me demonstrate.
>
> If the cap is not required and can be proven so by not proving otherwise,
we
> must have specified a set of logic levels that guarantees that there is
> little or no common mode offset between chips. Why not, then, specify a
cap
> and ask to have someone prove that the common mode offset specification is
> required? Can't prove it? Then we should reject a 0.75 common mode
voltage.
> Sound absurd? I agree.
>
> Please note, I am not taking a position on the choice. I don't feel
> qualified to do so. There are many reading this that are much more
qualified
> that I (we?).
>
> The point is that you win an argument by first establishing the
assumptions.
> I, for one, have not YET accepted yours.
>
> So, a few questions for those "in the know:"
> 1. What is the impact of this decision on parts that operate at different
> voltages and use different technologies (e.g. a "MAC Chip and a WDM
> Transceiver")? Yes, I reject the assumption that these could and should
use
> the same technology.
The decision I propose is to allow either AC or DC coupling as an
implementation decision not a standard mandate. Parts that operate at
different voltages and use different technologies would likely be
AC-coupled.
> 2. What is the impact of this decision on low jitter design?
I suspect that DC-coupling, if possible, is superior since the
interconnect is simpler.
> 3. What is the impact of this decision on EMI generation at the board
level?
Probably negligible.
> 4. What is the impact of this decision on signal integrity?
I suspect that DC-coupling, if possible, is superior since the
interconnect is simpler.
> 5. What is the impact of the ground differentials common between high
power
> boards in a system?
None. If a significant ground differential condition exists, use
AC-coupling.
> 6. What else am I forgetting to ask?
>
> jonathan
>
> > -----Original Message-----
> > From: Rich Taborek [mailto:rtaborek@xxxxxxxxxxxxx]
> > Sent: Friday, September 22, 2000 3:49 PM
> > To: HSSG
> > Subject: Re: XAUI AC coupling
> >
> >
> >
> > Jonathan,
> >
> > I can best describe your response as "twisted".
> >
> > XAUI is an optional interface for P802.3ae. I don't believe
> > that I will
> > get any disagreement on this point. However, 802.3 voters voted
> > unanimously to include XAUI as a baseline proposal in P802.3ae. The
> > definition of XAUI in the standard has been allocated a clause,
> > specifically, clause 47.
> >
> > We (P802.3ae) need to develop XAUI specs. The ones in the baseline
> > proposal are a good start but are far from complete. If you
> > were at the
> > New Orleans XAUI breakout session, and I believe you were for a short
> > time, you would have a good idea of how incomplete the XAUI
> > specs really
> > are. One of the issues at hand is one of Tx and Rx coupling method.
> > Recent reflector discussion is already talking about the possible
> > specification of capacitor values for XAUI AC-coupling.
> >
> > I'm taking my usual systems perspective and trying to have a serious
> > technical discussion on the issue of the requirement of a specific
> > coupling method for XAUI. I believe that this TECHNICAL discussion is
> > both appropriate and warranted. My point is that XAUI
> > AC-coupling is NOT
> > required for interoperability. Therefore, it should not be a
> > requirement
> > in the standard. Such a requirement, if any, would be listed
> > in the PICS
> > of clause 47. I don't believe that asking for proof of an AC-coupling
> > requirement for XAUI is extreme at all. This proof is in line with the
> > PAR 5 Criteria and the KISS principle, some of the basic
> > tenets by which
> > 802.3 has historically made decisions and made its standards so
> > successful.
> >
> > As further proof of the wishy-washy nature of the XAUI
> > "requirement" to
> > AC-couple I will submit a presentation by Ali Ghiasi on this issue:
> > http://grouper.ieee.org/groups/802/3/10G_study/public/nov99/gh
> > iasi_1_1199.pdf.
> > See slide 6 where the AC coupling caps are shown and accompanied by an
> > asterisk which reads:
> >
> > * It is recommended were possible to use 0.75V for the common mode
> > voltage
> > to allow the user for possible direct connect in backplane
> > applications.
> >
> > Well, fact of the matter is that XAUI is ONLY a backplane (i.e.
> > chip-to-chip) application. There is no objective in P802.3ae to extend
> > XAUI with copper cables between equipment which is likely powered by
> > different supplies. Therefore, even the original Hari presentations
> > suggest that DC-coupling is appropriate for XAUI.
> >
> > Lets please keep this discussion above board and on a
> > technical level so
> > we have a chance of finishing this standard on time.
> >
> > Best Regards,
> > Rich
> >
> > --
> >
> > Jonathan Thatcher wrote:
> > >
> > > Rich,
> > >
> > > While the discussion is worth having and your request for supporting
> > > information valid, I think that your conclusion that we
> > must prove that it
> > > is required for interoperability a bit on the extreme side.
> > There are lot's
> > > of things that we add to the standard, that have
> > significant value, that are
> > > not required for interoperability. For example: XAUI.
> > >
> > > jonathan
> > >
> > > > -----Original Message-----
> > > > From: Rich Taborek [mailto:rtaborek@xxxxxxxxxxxxx]
> > > > Sent: Friday, September 22, 2000 11:57 AM
> > > > To: HSSG
> > > > Subject: Re: XAUI AC coupling
> > > >
> > > >
> > > >
> > > > All,
> > > >
> > > > I'd like to propose that clause 49, XAUI, remove any
> > requirement for
> > > > coupling, AC or DC. The basis of this proposal is as follows:
> > > >
> > > > 1) XAUI is a chip-to-chip interconnect. As such, DC-coupling is
> > > > clearly appropriate and advantageous from an implementation
> > > > perspective, an example is when interfacing chips from like
> > > > logic families utilizing the same power supplies. This is likely
> > > > to be the case in many implementations. Therefore, the standard
> > > > should not dictate AC-coupling when DC-coupling is adequate to
> > > > achieve interoperability.
> > > >
> > > > 2) I've reviewed all instances of the use "coupling"
> > including fuzzy
> > > > variants in the 802.3 standard. There is no precedent for
> > dictating a
> > > > specific coupling method for a chip-to-chip interconnect in the
> > > > standard.
> > > >
> > > > 3) Absolutely nothing will be taken away from the standard by
> > > > removing a requirement for AC-coupling. If AC-coupling is either
> > > > desired when not required or required for a specific
> > implementation,
> > > > then the details for AC-coupling including the determination of
> > > > specific capacitor values, the frequency spectrum of 8B/10B
> > > > transmission code, etc. are all well documented and
> > readily available.
> > > > 8B/10B transmission code is far and away the most
> > commonly used and
> > > > well understood code in serial gigabit links including
> > chip-to-chip
> > > > interconnects. From a signal coupling perspective, except for
> > > > proportionally higher signaling frequency, there is no difference
> > > > between a single XAUI lane and a 1000BASE-X link. Note that for
> > > > 1000BASE-X, both AC and DC coupling is available from
> > > > transceiver module vendors.
> > > >
> > > > AC coupling was proposed as a requirement for the Hari
> > interface which
> > > > was effectively renamed as XAUI. It has been carried into
> > the baseline
> > > > proposals for P802.3ae. Now is the time to decide whether
> > AC-coupling
> > > > is an interoperability REQUIREMENT. I challenge anyone to
> > argue and
> > > > prove that AC-coupling is required for XAUI
> > interoperability. If such
> > > > proof is not forthcoming, clause 49 should be modified to
> > remove any
> > > > requirement for AC-coupling.
> >
> > -------------------------------------------------------
> > 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
> >
--
Best Regards,
Rich
-------------------------------------------------------
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