| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Can we use systems that folks
designed to KR channel requirements? Which were in essence based on Joel’s
framework. It’s not same situation.
The server market is presently high growth. So the installed base plays a
bigger role now than then. … Rich From: John D'Ambrosia
[mailto:jdambrosia@xxxxxxxxxxxxxxx] All, First, going into specific
materials would be going into implementation. That is left to the
implementer. So 802.3ap had a similar
problem, and eventually a set of properties (associated with materials thought
to be candidate materials for use) was used to define “improved
FR-4”. (Thanks Joel!) Here is one link I found - http://www.ieee802.org/3/ap/public/may04/goergen_01_0504.pdf.
These values were adopted at the May 2004 meeting (see http://www.ieee802.org/3/ap/public/may04/minutes_1_0504.pdf).
I believe they were later updated, but you can get the idea from these URLs. It sounds like having a
definition for whatever we call the board material (improved FR-4, enhanced
FR-4, etc) could be useful, and I would look to participants in the SG to start
considering this. In the meantime I will go through the 802.3ap archives
and see what the final agreed upon definition of “improved FR-4”
was for 802.3ap. Regards, John From: Chalupsky, David
[mailto:david.chalupsky@xxxxxxxxx] At the last meeting we discussed briefly what is the
“legacy” channels we should attempt to cover. Obviously it’s not a 10GBASE-KR channel – not enough
lanes. As has been suggested by others, I agree that we should look at
40GBASE-KR4 channels as the ‘legacy’ we should attempt to
support… especially for the bladeserver market. Kinda funny,
because KR4 is not common in the market yet, but will be soon… so
it’s a “future legacy.” As Rich points out PCB technology moves at a snail’s pace
compared to silicon. I started designing boards in 1988 and the
technology – especially for volume PCs & servers - is largely
unchanged. In fact, your Smart Phone uses much more advanced PCB tech
than an IA server. Look at the size of a Server
motherboard – if we force a technology that doubles (or more) the
square-inch cost of the PCB, that is not a good fit for this market.
Relating this to the objectives terminology below, if “improved
FR4” means Megtron-6, that is impractical. We have another external force is this market: regulatory
requirements for RoHS (restriction of hazardous substances).
Over the last few years the PC/Server market made the
transition to lead-free PCBs in support of RoHS. Our next exciting
challenge is Halogen Free. That includes getting rid of the bromine &
chlorine flame retardant common in FR4 today. This makes a fundamental
change in the epoxy dielectric properties. (Actually the transition
to lead-free was easier electrically – it was primarily a change in
surface finish away from HASL; didn’t affect the dielectric.) There
is active industry effort underway to improve the electrical performance of HF
PCBs… but the first stop along that path is just getting them up to
“common” FR4 performance. Finding a cost-effective,
manufacturable, Halogen Free PCB solution with the electrical performance of
Megtron-6 (or -4, or Nelco-13SI) may be a long way off… but more
investigation is needed here. For more background on HF, see: http://thor.inemi.org/webdownload/projects/ELSC/HFR-Free_PVC-Free_Timeline.pdf Tying this back to a length objective for backplane channels is
difficult, but I think we had the same issue in 802.3ap. We assumed
“improved” FR4 to achieve 1m. But in practice I think most
implementers chose to stay with lower cost PCB material & implement shorter
channels. In 802.3 we are used to having an objective for length –
but that’s an awkward fit. Dave From: Mellitz, Richard
[mailto:richard.mellitz@xxxxxxxxx] How about dielectric material
that is projected to be used in X percent of the market in 201#? That
plays to broad market potential. Silicon technology is expected to still have
exponential growth. Changing “backplane/line card” board
technology on the other hand is like turning a battle ship. Given
that, we should at least evaluate how much of legacy product present in
201# we can cover given the disparity of the silicon vs. board technology pace.
Simply put, coverage could be one of the metric to compare proposals.
Unfortunately it’s still crystal ball. … Rich From: Havermann, Gert
[mailto:Gert.Havermann@xxxxxxxxxxx] All, besides the grammar I would like to have some kind
of reference to the "enhanced FR-4 Material". There are so many
different anhanced materials available (low loss, ultra low loss, low dk
glass...) all having a different influence on the length factor we are trying
to fix (not to mention the cost differenve between materials). Regards Gert Havermann Absender ist HARTING Electronics GmbH & Co. KG; Sitz der
Gesellschaft: Espelkamp; Registergericht: Bad Oeynhausen; Register-Nr.: HRA 5596;
persönlich haftende Gesellschafterin: HARTING Electronics Management GmbH; Sitz
der Komplementär-GmbH: Espelkamp; Registergericht der Komplementär-GmbH: Bad
Oeynhausen; Register-Nr. der Komplementär-GmbH: HRB 8808; Geschäftsführer:
Edgar-Peter Duening, Torsten Ratzmann, Dr. Alexander Rost Von: John
D'Ambrosia [mailto:jdambrosia@xxxxxxxxxxxxxxx] All, Not sure this message got
forwarded to the reflector, so forwarding. Regards, John From: Kolesar, Paul [mailto:PKOLESAR@xxxxxxxxxxxx] John and group, The objectives are on the right track, but there are a few
grammatical and clarity issues. The first is an issue with the wording at the end of each.
The phrase “up to at least” is illogical and confusing
because it combines words that specify maxima (i.e. up to) with words that
specify minima (i.e. at least). While this phraseology may be following
some precedent of former objectives, it only serves to cloud the real intent.
In the past these objectives have always been interpreted as the
requirements for minimum reach. Therefore I propose that they each simply
state it as such by replacing “lengths up to at least” with
“lengths of at least”. Further, I do not know what is intended by the phrase “for
links consistent with lengths”. Why use the word
“consistent”? Is there a perception of some increased
flexibility or some other advantage? Please explain. If the
advantage is ambiguity, I would prefer stating objectives more crisply. Unless your rationale for this word choice is compelling, I propose
combining these two issues into the following new phrase: “for link
lengths of at least”. Lastly, it would be simpler and clearer to place all the adjectives
describing “traces” before the noun rather than some before and
some after. Rearranging these yields ”over improved FR-4 copper
traces”. With all three of these changes, the two draft objectives become: · Define a 4-lane 100 Gb/s PHY for operation over improved
FR-4 copper traces for links lengths of at least “X” m. · Define a 4-lane 100 Gb/s PHY for operation over copper
twin-axial cables for links lengths of at least “Y” m. Of course all these grammatical improvements do not address the
main issues which are the values of X and Y. But that is what study
groups are for. Regards, Paul Kolesar From: John D'Ambrosia [mailto:jdambrosia@xxxxxxxxxxxxxxx] All, I wanted to try and foster some
discussion on the reflector regarding objectives for the project to help all focus
their planning of presentations for March. So what do we appear to have
consensus on so far? a)
We are in a study group looking at
100GbE over backplane and copper twin-ax b)
Legacy support indicates broad market
potential would be aided by 4 lane solutions What appears to need further
consensus building? Well the big ones would seem to be reach for both
backplane and cu cabling objectives. So if we can combine where we
appear to have consensus with what we need to resolve, the following two
statements could be used as strawmans for objectives for the group to work
towards (leaving the reach #’s as variables for now): · Define a 4-lane 100 Gb/s PHY for operation over copper
traces on improved FR-4 for links consistent with lengths up to at least
“X” m. · Define a 4-lane 100 Gb/s PHY for operation over copper
twin-axial cables for links consistent with lengths up to at least
“Y” m. Feedback? John |