Test Point Discussion - Revived
IEEE P802.3ap Task Force Members,
I apologize for not being able to join
that channel model ad hoc call this morning. It is my understanding
that we continue to have disagreement on the placement of test points
for the purpose of transmitter, channel, and receiver definition.
I cannot over-emphasize the need to
find the common ground on this issue, put a stake in the ground, and
move on to other pressing topics.
After observing the discussion on the
reflector, I tried the summarize the salient points of each principal's
argument and attempted to find the common ground between them. This
summary is in the reflector message (http://ieee802.org/3/bladesg/email/msg00166.html) dated the 26th of August. I received no
feeback on this summary but clearly there are still issues with this
approach.
I have copied the entirety of the
summary message for easy reference. I strongly encourage those with
reservations regarding this proposed approach to state those
reseverations clearly and offer a proposal for resolution that can be
evaluated by the Task Force.
Finally, I hope it is clear the
proposal below is not the position of the Task Force chair nor is it
necessarily my personal preference. It is simply my perception of the
common ground which I have summarized as an aid to generating
concensus. Please, take the time to review it carefully, provide
constructive feedback, and help me find the common ground so that we
may move on.
Thank you,
Adam
Title: Test Point Discussion
Summary
IEEE P802.3ap Task Force Members,
I am just now catching up on the test
point discussion. My eyes may be deceiving me, but after carefully
reading the thread, it appears that we may be close to consensus on
this. Let my try to summarize what I saw as the salient points of this
discussion and see if I can flush out any remaining points of
disagreement.
1. The convention used for the
DEFINITION, SIMULATION, and SPECIFICATION of the transmitter, channel,
and receiver, is that the AC-coupling capacitor is between TP4 and the
receiver.
1a. Transmitter specifications are
defined looking into the transmitter from TP1.
1b. Receiver specifications are
defined looking into the receiver from TP4.
1c. Channel model ad hoc defines
the signal path from TP1 to TP4.
Discussion:
I believe this is consistent with Mike
Altmann's observations about how devices are actually tested by third
parties and during interoperability plug-fests. The observations points
in such environments are the test points provided on the evaluation
board, and that typically includes some type of connector (SMA), some
distance of PCB trace, the landing pad, and on the receiver side,
typically an AC-coupling capacitor and associated mounting structure.
These are connected to "test" backplanes, examples of which we have
seen presented by Joel Goergen, John D'Ambrosia, David McCallum, and
Steve Anderson. Under this convention, we are still able to use such
environments to demonstrate feasibility and interoperability.
2. We define the signal path from the
transmitter to TP1, and from the receiver to TP4 to be as "transparent"
as possible.
2a. Of course we realize that these
paths will not be completely transparent. To borrow a phrase from Rich
Mellitz, we need to define the signal paths "whose error terms are
acceptable to all." It is my recommendation that these signal paths are
defined as part of the work of the signaling ad hoc and that we use
Rich's "super-cap" model as the basis for this work.
2b. Transmitter and receiver
electrical specifications as defined at TP1 and TP4 take the reference
structures defined per (2a) into account to avoid penalizing the
components for any lack of "transparency".
2c. All performance simulations
include these reference structures.
Discussion:
I believe that this is consistent with
the recommendations made by Rich Mellitz, John D'Ambrosia, Joel
Goergen, and Charles Moore (option #2).
3. Channel specifications span TP1 to
TP4. The specifications include a footnote or an appendix describing
the environment defined in (2) above for which the specifications were
verified (equivalent circuit, return loss/insertion loss curves for
Tx-to-TP1 and TP4-to-Rx, etc.).
Discussion:
As Mike Altmann points out, TP1 and TP4
are the demarcation points for the standard, and we need to ensure that
they "match up" in order to guarantee interoperability. Per the
statements above, ownership of the AC-coupling cap and associated
mounting structure to the receiver "block". However, as Rich and others
have observed, per the system designers choice, the channel may not
include the AC-coupling cap or utilize different placement or a lower
cost/performance alternative to what we define above. The tricky part
is making it clear that if you "cheapen up" on the AC-coupling
structure, you need more margin from TP1 to TP4 to make up for that.
Adding informative statements related to the specification assumptions
for Tx-to-TP1 and TP4-to-Rx are one way to address this as the allow to
interested system designer to understand the trade-off. I say
"informative" because making such statements normative implies that we
are mandating a particular AC-coupling structure, which is not what we
want to do.
I think this sums it up. If I
misquoted any of you, I apologize and please correct me as appropriate.
If any of you have a fundamental disagreement with any the above
points, please air it out now so that we can address it. This Task
Force has a great deal of work to do and schedule to keep. The time has
come to put a stake in the ground on this and put it behind us.