****************************************************** Minutes of the IEEE 802.3z Gigabit Ethernet Task Force Irvine, California, March 10-14, 1997 ****************************************************** Note: I tried to correctly attribute comments to each speaker, but there are many faces I couldn't immediately match to a name. Also, for the sake of typing, I have abbreviated some of the common speakers by their initials, as follows: HF=Howard Frazier HJ=Howard Johnson RT=Rich Taborek PT=Pat Thaler JT=Jonathan Thatcher SS=Steve Swanson MM=Mart Molle SH=Steve Haddock SM=Shimon Mueller RS=Rich Siefert DL=David Law ****************************************************** The meeting was brought to order Monday morning by the Chair, Howard Frazier. The proposed agenda for the meeting was distributed. 1. Introductions **************** Each attendee stated their name and corporate affiliation. There we approximately 150 attendees at the opening of the meeting. 2. Minutes Secretary ******************** The Chair volunteered Bruce LaVigne to record the minutes of this meeting. 2a. Review Agenda ***************** The Chair went through the proposed agenda, including the presentations to be given in each of the four tracks. It was noted that the contents of the tracks had been slightly altered from previous meetings: since the track I (system) presentations were many, especially with the addition of management, it was split into two. The tracks for this meeting are: Track I - System/Repeater/MAC/Management Track II - GMII/PCS Track III - PMA/Short Haul Copper PMD/Optical PMD Track IV - 1000Base-T The Chair then asked if there were any changes to the agenda. There were only minor changes: Track I: Add Mart Molle(UC Riverside): 15 mins - MAC Issues for Burst Timers. David Law only needs 60 minutes, not 120, to discuss Management Issues. Track III: Add Steve Swanson(Corning): 15 mins - Review Optical Fiber References. Track IV: Removed presentation 6 which HF had thought Salesh would want to give. A motion to approve the agenda was then entertained. Moved: Bill Quackenbush Seconded: John Payne Passed by acclamation 3. Ground Rules For Connector Discussions ***************************************** In the interest of letting everyone hear and participate in the discussion relating to connectors in an orderly manner, the Chair recommended the following ground rules: i. Rich Taborek, FC liason, to give status of recent X3T11 actions with regards to connectors. ii. Connector presentations given to entire 802.3z on Wednesday morning for any & all types. iii. Motions for connectors to be made on Wednesday afternoon, not in any sub task force meeting. iv. If no motions pass, continue to spec duplex SC connector for fiber optics, and DB-9 for short copper. 4. Status Report **************** The Chair gave an update of the current position of work, including the fact that Draft 2 was distributed on 24-Feb-1997. He also reviewed the .3z voting rules (vote if present & qualified, 50% procedural, 75% technical) 5. Timeline *********** The Chair reviewed the timeline graph, noting that we were at the halfway point to a March, 1998 Standard, with lots of hard work in the months ahead. 6. Email Reflector/Web Site *************************** The Chair noted that there were now approximately 660 names on the reflector. He provided detailed information for subscribing and sending to the email reflector. The web/ftp site was also reviewed. In addition, since the Chair is expecting approximately 2000 comments on the draft, a web comment form has been set up, and an ASCII version will be posted. The Chair would like to, but doesn't expect to use electronic balloting. HJ would like to see company name field on web comment form (optional). No one objected to this. Bob Fink reminded folks that one comment per web form is intended. People wishing to submit long, involved comments, or a multitude of comments, will probably want to use the ASCII version. 7. Review HSSG Objectives ************************* The Chair showed the 15 objectives, and reminded everyone that we must either work to these objectives or vote to change them (which is OK, but it must be done). He mentioned the 25 meter limit on short haul copper as a possible example, and JT spoke up saying they hoped to meet this. 8. Distribution of Documents **************************** Additional copies of Draft 2 were handed out to anyone who did not already have one. The Chair was happy that he had not killed trees in vain ;-) HJ gave an overview of Draft 2. Highlights: updates from January meeting, autonegotiation from clause 28 into clause 37, a new management clause 30. 9. Call for Patents ******************* The Chair issued a call for patents. He will provide a sample form letter and copy of the draft to anyone who needs it. A complete set of patent letters is on file at IEEE offices. Q: has anyone already signed up? A: only IBM for 8B10B coding so far. 10. Liason Reports ****************** TIA TR41.8.1: Chris Di Minico Enhanced Twisted Pair Cabling Systems PN 3727 draft available. Their next meeting is mid April. Patch Cord & connecting hw test measurements ScTP (100 ohm screened twisted pair) draft12 letter ballot PN3193 Outstanding ballots: 3772: prop delay/skew (approved) 3723: re-issued mods Fibre Channel X3T11: Rich Taborek New name: NCITS (check out www.x3.org or www.ncits.org) New fibre connector selection process: pared down to 3 connectors for FC-PH. Galaxy connector by 3M & Honeywell passed after lots of motions/ammendments. URL to see minutes: http://www.amdahl.com/ext/CARP/FCA/FCA.html Steve Joiner noted that there is no formal proposal for project using this new connector. 11. Review Rooms **************** The Chair discussed decision making: if a breakout group can resolve an issue without involving other groups, please do so! Thus, subtask force groups can have motions, straw polls, etc. Overall issues are left to motion madness on Wednesday afternoon. Also, please work to fill in TBDs! 12. Sub Task Force Breakouts **************************** Track I: System/Repeater/MAC/Management 12.1.2.2. Management Issues [David was ready before the others] --------------------------- David Law (Cisco) Overview of the new management clause 30: SubTaskForce formed Jan 97; first draft of cls 30 in D2; Sumesh Kaul is the editor. This week: line by line, review issue list 12.1.1.1. Draft 2 Changes in Repeater (clause 41, 42) ----------------------------------------------------- Steve Haddock (Extreme) Clause 41 mods (D1->D2): Convert ByteTime to BitTime Rename FCCLimit, CCLimit to FCELimit, CELimit (Count to Event) Received Event based on CRS with CRS based only on receive activity Repeater Core state machine changes: -add preamble and wait states (but now remove wait for D2.1) -split error wait out of jam -use CRS and COL inputs (but now remove COL for D2.1) Partition State machine yet to be distributed Text on Grounding/Safety issues CRS issue: SH wants to motion on Wednesday that PCS/GMII loopback be disabled for repeater so CRS can be only rx, not tx. Remove wait state from repeater state machine. Other state machine mods: use CRS instead of COL for faster response, transition from preamble to idle if CRS goes away (to prevent possible lockup in preamble) A motion to accept the clause 41 modifications was made. Moved: Richard Grenier Second: Walt Thirion Passed: unanimous (with many not voting) Clause 42 mods (D1->D2): Remove Buffered Distributor Add TW-style cable to tables Recalc max cable lengths with no margin Will include full duplex link table Remove references to inter-repeater links Convert ByteTimes to BitTimes HF asked to add footnote to cable length table that 1000BT (CatV) is work still in progress. HF asks what should we do about all these new (better?) cable types coming out? Answer: they are not in the standard. 12.1.1.2. Burst Timer Lengths' Simulation ----------------------------------------- Mohan Kalkunte (AMD) Investigate impact of increasing burst timer length on network performance -Uniform loading, non-uniform loading of stations -packet end-to-end delay, access latency Questions were asked about higher offered loads (slides only go to 60%); Mohan replied that those sims were presented at earlier meetings(12K only) Summary: significant improvement from 0 to 12Kbits, some to 24Kbits, minimal going to 48Kbits. Capture effect does not increase with larger burst size, just opposite (since less chances for collisions). HF raised a concern about other changes (excessive defer?) to be made if we raise the limit above 12Kbits. 12.1.1.2.5. Making Half-Duplex Repeaters Attractive for Gigabit CSMA/CD ----------------------------------------------------------------------- Mart Molle (UC Riverside) BEB (normal backoff) isn't affected much by burstLimit; but if BLAM gets adopted, it is *VERY* suceptible to the increased slot time - so current 12Kbit burst w/BLAM maxes out at 30-40% load. However, with 8Kbit burstLimit, BLAM looks better than BEB out to ~80% offered load. Also, Mart is recommending that repeaters truncate collisions - this helps by ~2x. Questions were raised as to what happens at higher load? Mart's answer: anything is better than capture effect - which gives a station around 100 packets in a row for gigabit links. SH noted that since this is a timer value, it could also be changed as late as July; or perhaps a motion today/tomorrow, or comments to D2. 12.11..4. D2 changes in Clauses 1, 2, 34 [note: HJ & HF swapped] ---------------------------------------- Howard Johnson (Signal Consulting/Packet Engines) Misc minor changes (see D2). A question was raised about table 34-1: what is correct fibre size 9um or 10um? I didn't hear the answer. 12.1.1.3. D2 changes in Clause 3, 4 ----------------------------------- Howard Frazier (Cisco Systems) Picture, text, describing extension field as an (optional) part of the frame added to clause 4. Diagrams, text, describing packet bursting, length of carrier event added to clause 4. Andy Luque began discussion about naming BurstLength; change to BurstLimit SM expressed the desire to see draft-to-draft changes; it was decided to use an editorial block at beginning of clause describing them. Long discussion of collision filtering & late collisions at receiver Scott Mason suggested we: introduce late collisions, receive criteria for late collisions, state that MAC recieves frames frame-by-frame, and describe conditions of collision hitting frame or IPG. HF then went on to describe the functions of "PhysicalBit" - tx can have 4 values (0, 1, extend, extend_error), rx only 3 vals (no extend_error). He then delved into the changes to the Pascal code: TransmitLinkManagement: made burstTimer an independent process, so just look at the variable "bursting"; late collision jams with car_ext_error, tx not retransmit on late collision. SM discussed getting rid of extendSize in favor of using slotTime/minLength formulas. A long discussion about frameWaiting and bursting followed; SM has possible revisions... Move increment of interFrameCount above collision detect so really send 32 bits of JAM (PT pointed out that especially during carrier extend, this is not that important). ReceiveFrame: is it a function or procedure? Check with RS on this. Rich shows up! The answer is: no particular reason - so change it in the description to match the function. Another action: look at daggers- layer management lines in pascal; also look at clause 5: since it was deprecated, the pascal in clause 5 has "gone away" but is necessary. SM is concerned that as written, the pascal may allow alignment errors when !extensionOK. ReceiveLinkMgmt: now part of collision filtering is done in bit receiver -- the bursting check is done there since it knows which packet in a burst it is. StartRecieve: since we are deleting the initialization of currentReceiveBit, should move this definition into the bit receiver as a local var, from the global definition on 4.6. BitReceiver: whole procedure has changed. Discussion about enableBitReceiver and the asynchronous nature of the pascal procedures. Clarification: receiveDataValid is NOT THE SAME as RX_DV on GMII. It is derived from the PLS service primitives. In particular, receiveDataValid is asserted for entire burst, whereas RX_DV drops between packets! MM: put an "else" in at line 27.5 to show that you don't need to add data bits to frame after you have gotten some extension. No one cares, so we'll leave it as is. SM, RS, et. al. discussed pascal style comments on 32, 39. Line 36 needs to use enableBitReciever instead of "if receiving ..." for consistency. SM wants to get rid of extendSize parameter by using slotTime instead. One concern is why there were two parameters to begin with; another one is clarity: sometimes we are really looking at extensions as opposed to collision slot times... SetExtending: new independent process, sets "extending" variable when in half-duplex gigabit mode in between packets. Note: move same construct to the Initialize() procedure so other media still OK. Issue with Pascal: burstMode is a variable, but currently with no way to set it. SM would like to use burstMode for all carrier extend activity. Others disagree that then turning it off would disable receiving carrier extensions. Page 4.21 add text to ReceiveDataValid to point to clause 6 so people don't think it is identical to RX_DV. Parameter tables: one table for each speed. extendSize set to 0 bits for all speeds below Gigabit. Actually, move it out of table to a constant, which is set to (slotTime - minFrameSize). That way, can use both in pascal, to show what is intended at each point, but only slotTime needs to be in table. burstLength (now burstLimit) currently 12000 in all tables; can we change to "Not Applicable" for non-gigabit tables? We think so. MM wants to extend burstLimit to make it easier for BLAM (if it happens) Geoff says it puts us on the slippery slope to bumping up frame size Large amount of discussion on this. Resolution: Mohan to do some more sims on bigger bursts, higher offered loads, for the next meeting. DL is concerned over IPG shrinkage. SH says it can be up to 4 bytes with gigabit's single repeater. We should put a warning in the standard, noting that transmitters CANNOT use smaller IPGs but receivers may see them. 12.1.2.1. Review of Clause 30 (Management) ------------------------------------------ Sumesh Kaul (Bay Networks) Draft 2.1 will include changes for .3x and .3y as well. Not done for D2. SH tasked with reviewing repeater functions, even though we ignored it at 100T Currently add 3 optional counters for Gigabit repeaters: aBursts, aBurstsXmitted, aBurstsReceived. Change name of column to Burst Monitor Capability so others may use it later. Change name of DTE 100Mb column to Symbol Monitor Capability (since this is the only counter represented, and thus gigabit uses it too). Since it is only on repeaters, no reason to have all three burst counters. Drop 2 objects and only support aBursts (received). Count once for each carrier event (including single, non-bursted packets). Discussion about clause 5 being deprecated - too much (still need pascal). Also, we will remove editor's notes for clause 5 on each counter in favor of one macro comment. HF suggested we add aBurstCapable to .3x register aMACCapabilities (not in D2) Also, add get/set aBurstModeEnable attribute only for half-duplex gigabit. Make sure that collision counters reference slotTimes, not 512BT. Deprecate aOutOfRangeLengthField counter. aFrameTooLongErrors - deal with 802.1Q increasing frame size: use text in clause 4.4.2 of .3x, make sure that it is discussed at the plenary level so everyone is clear on what we do. Must do global search/evaluate for LLC to make sure that we mean LLC and not MAC client. For aPhyType, delete "none" since can't have removable GMII. Footnote for 1000T: a specification for 1000BaseT is plan for future work. Decide to keep 1000BaseT placeholders in clause 30 for now, may remove them if they are too far behind in May/July. aMIIDetect for gigabit, return the value "absent". For aPhyAdminState, add text for GMII pointing to clause 35.2.1.5. For aTransmitCollisions, etc., get repeater state names from SH just before Working Group ballot. Add note for alignmentErrors indicating gigabit can't have these. (some want it deprecated). aShortEvents: use a value of 72 BT so we stay away from the 96 BT limit. aRunts: this counter is counting collisions in another repeater; since gigabit has only one repeater, this counter is meaningless, so we dump it ;-) LateEvents: definition seems to be useless, and inaccurate, esp. for gigabit. Lots of discussion about what it might have meant at 10, 100, and what would be useful to customers. Open issue, note to reflector. aDataRateMismatches: counts overruns/underruns; wording suggested which I missed (check with Sumesh, or see the next draft). aBursts: define a burst as a carrier event at least a slotTime long. aMediaAvailable: 2 choices: tell PMD group that the one place to use signal detect is aMediaAvailable; or get all info from PCS, so don't need signal detect from PMD at all. aJabber: not supported nJabber: not sent On page 30A.45, fix enumerated list: only 1000X (no distinction on phy), so: X=36, XFD=362, T=40, TFD=402 On page 30A.47, same holds, so: LX=38, SX=382 ------------ end of Track I -------------- Minutes 802.3z GMII Subtask Force Monday Mr. Bob Grow conducted the portion of track 2 covering clause 35, Gigabit Media Independent Interface. The changes made from 802.3z/D1 to 802.3z/D2 were summarized. The first area of technical change was to accommodate the changes made to clause 4 for generation of jam during carrier extension. This produced changes in the mapping of PLS service primitives sections and in text describing error handing. The second area of technical change was to adapt the clause 22 management register set for gigabit operation. This was done by defining one control bit and one status bit in the base register for 1000 Mb/s operation. Definitions of existing bits applicable to 1000 Mb/s operation were edited, and gigabit specific bits were defined and assigned to reserved registers. The 802.3z/D1 comments were mostly editorial. The comments were reviewed and open issues highlighted. Most of the comments left unresolved are in the electrical interface section. Mr. Grow expressed concern about the lack of progress since the January meeting in defining an electrical interface. He expressed the alternatives he saw for the subtask force and suggested that the 10-bit electrical interface be heavily leveraged to enable timely specification of the GMII electrical interface. This would include adoption of both the clocking and signal characteristics of the interface. There was significant opposition to this approach. Discussion and a straw poll indicated continued support for definition of a GMII electrical interface. To accelerate resolution of problems, some basic objectives were established for a group of experts to define the basic characteristics. This group was tasked with finding an approach to the electrical interface that could be included within 802.3z/D3, and to report progress Tuesday afternoon. Comments were solicited during a page by page review of the draft standard, as well as confirmation of basic directions taken in implementation of D1 changes. An open comment from the first draft asked if changes were required for clause 6. 802.3x modified figure 6-1 to be based on figure 1-1. Mr. Grow indicated that the issue had been discussed with other editors and presented Mr. Howie Johnson's proposed resolution based on those discussions. The proposal recognizes that clause 22 and clause 35 each redefine the PLS service primitives, and therefore, modification of the figure is appropriate to conform to the new figure 1-1 and illustrate the scope of clause 6. Mr. Brown moved and Mr. Dineen seconded a motion to: Adopt "Proposed Changes to Clause 6". The motion passed Yes: 22, No: 0, Abstain: 3. The GMII subtask force recessed to Tuesday afternoon to hear the report of the GMII electrical experts meeting. Tuesday Tuesday afternoon, Mr. Dineen presented the report of the recommendations to the subtask force. A similar presentation had been made earlier to the 1000BASE-T subtask force and was well received there. At the conclusion of the presentation, Mr. Dineen moved and Mr. Fifield seconded a motion. "Move to accept the GMII electrical subtask force's report for inclusion in the next draft of the standard (draft 2.1)." The motion passed Yes: 24, No: 6, Abstain: 6. REPORT TO 802.3z Clause 35 - D1 to D2 o Clause 4 consistency o Incorporation of clause 22 management registers o Many editorial changes o Open comments are primarily on electrical, no progress made on electrical from January Clause 35 - D2 review o Continued support for electrical interface - Small group of experts generated recommendations - Subtask group vote to include recommendations in next draft o Plan in place to make register set consistent with clause 37 o Motion for clause 6 change page adopted ************************************************************************* Track II - PCS, PMA, LC Report - Clauses 36 and 37, which include PCS, PMA, and LC, and comments and responses to those clauses, were reviewed in Track II. o Session held on Tuesday, March 11, 1997 from 1:30 p.m. until 9:30 p.m. with breaks - Reviewed Draft 1 comments and responses including 154 PCS, 46 PMA, and 13 LC comments. o All open D1 comments not resolved in D2 are included in the D2 comments and responses document. o D1 comments and responses document uploaded to IEEE FTP site. Eventual location should be: ftp://stdsbbs.ieee.org/pub/802_main/802.3/gigabit/presentations/ mar1997/RTd1com.pdf - Reviewed all current Draft 2 comments and responses to date. o D2 comments and responses to date include 58 PCS, 11 PMA, and 14 LC comments. o Many D2 comments were resolved. o Directions established for resolving most of the remaining comments. o D2 comment period is open until the cutoff date for the next draft. o Please submit D2 comments via the IEEE 802.3z reflector: stds-802-3-hssg@mail.ieee.org o D2 comments and responses draft (D2 comments submitted before March 10, 1997) uploaded to IEEE FTP site. Eventual location should be: ftp://stdsbbs.ieee.org/pub/802_main/802.3/gigabit/presentations/ mar1997/RTd2com_draft.pdf o Ben Brown of Cabletron presented changes to PCS Transmit, Receive, and Synchronization state machines based on proposed resolutions to D1 and D2 comments. + State machines available in RTd2com_draft.pdf. + Additional comments received, primarily Òcorner casesÓ. + Basic state machine operation appears to be solid. + Call for simulation of state machines. - Clause 36 (PCS, PMA) Work List: o Primary differences between Clause 36 D2 and D2 comment resolution are changes to state machines. o All existing D2 comments to Clause 36 can realistically be resolved and Clause 36 updated within a week. o Clause 36 ready for ÒearlyÓ draft revision to enable thorough review at the May meeting. - Clause 37 (LC) Work List: o Substantial changes required to Clause 37 to resolve nomenclature differences with Clause 36. o Several technical changes required including: + Link break; + Addition of State machine timer/count for each handshake; + Management register definitions; + Delete support for multiple PMAÕs; + Pause clarification. o Resolution to all outstanding Clause 37 comments by the editor will take several weeks. o Clause 37 NOT ready for ÒearlyÓ draft revision. o Desire to rev Clause 37 in time for review in May. ************************************************************************* Minutes from IEEE 802.3z plenary meeting held in Irvine, CA on March 10 - 12. Track III - Optical and Short Haul Copper SubTask Force The optical and short haul copper meeting commenced on Tuesday March 11 with Jonathon Thatcher reviewing the previous day's selection of team leaders for the newly designated "sub sub task forces" or teams. The team leaders were identified as follows: 1000 Base CX - Ed Grivna (Cypress Semiconductor) Copper Cable - Christine Foster (W.L. Gore) Shortwave & longwave optics - Del Hanson (HP) Optical Cable - Steve Swanson (Corning) Upcoming meetings and agendas for these teams were identified as follows: 4/11 - Friday of ANSI Fibre Channel standards meeting, Palm Springs, CA AGENDA: * Identify and enlist team members * Review Draft 2, anticipate written comments * Review itemized set of action items with owners 4/24 - Biltmore Hotel, San Jose, CA AGENDA: * Review recommendations created from Palm Springs meeting * Create motions (technical) and direction (editorial) for May interim meeting The agenda for the meeting in progress was presented by Jonathon. A motion was made by Stan Swirhun (Vixel) and seconded by Del Hanson to approve this agenda. This motion passed unanimously. Sign-up sheets were circulated for each of the task teams. A separate reflector will be set up and administered by Schelto van Doorn (Siemens). This reflector will serve as a data base for PMD related comments. Jonathon has agreed to work with Rich Taborek to create a PMA team whose work will encompass the GMII interface and the PMA to PMD interface. They are looking to identify a team leader, enlist members, and come up with a work plan. Jonathon has agreed to work with David Law to determine the requirements of the management layer. Jonathon is to bring back recommendations to the Track III group. Also, David Law will be invited to give an overview of this work at a date TBD. At this point, work items for each of the individual teams were reviewed as follows: OPTICAL CABLE TEAM (Work Items) * Change optical cable specs. to reference existing cable plant specifications. * Restructure clauses related to the fiber spec. to be based on fiber type, not wavelength. * Comments were received about references to 9 micron fiber, previous documents use 10 micron. * 38.10.3 - per last meeting, need to remove slope template and include specific values. * 38.11.2 - do we need a pull strength spec.? Is it in IEC 874-14? Motion - To accept the slide as presented by J. Thatcher as the basis for the working group. Moved - Schelto van Doorn 2nd - Peter Pepeljugoski (IBM) This motion passed with all in favor except for 2 abstentions. ******************************************************************************* 1000 BASE - CX (Work Items) * Comment received on differential voltage, desire to make compatible with future technologies. * Comments received regarding ESD spec. * Normalized amplitude does not match differential amplitude, needs to be resolved. * Fix TBD on distance spec. * General - measurement techniques subsection. * 39.6 - Wording requires some changes, is this section complete? ******************************************************************************* COPPER CABLE (Work Items) * Comments were received concerning desire to tighten certain specifications. * Comments were received regarding global use of terminology. * Comments were received regarding mechanical specification of Cu cable. * Need to close on spec. for measurement of connector loss vs frequency ******************************************************************************* SHORTWAVE & LONGWAVE OPTICS (Work Items) * Need to remove TBDs on link lengths. * 38.3.3 - Expecting input from J. Thatcher & D. Law to help define requirements for chatter, delay, etc. * Need to reolve Relative Intensity Noise spec. * Editor needs input on max. allowed optical power vs wavelength. * Need to define conditions for receiver sensitivity measurements. * Jitter budget allocation needs further definition. * 38.6.6.3 - Could this information be incorporated in the link budget table? ******************************************************************************* At this point, Jonathon asked if there were any volunteers for the position of vice-chairman of the Optical/Short Haul Copper Task Force. Jonathon explained that there may be instances when he will be unable to be present at the meetings. A motion was made by Howie Johnson to elect such a vice-chairman. This motion was unanimously approved. Schelto van Doorn was nominated by Stan Swirhun and seconded by Ed Grivna. Stan Swirhun was nominated by Howie Johnson and seconded by Schelto van Doorn. Del Hanson was nominated by Bill Sorels and seconded by ?. At this point, Stan Swirhun withdrew his nomination. A vote was taken to elect the vice-chairman. the results of the vote were as follows: Del Hanson 13 Schelto 27 Abstain 9 At this point in the meeting, editorial comments were reviewed to be applied to Clause 38/39 of Draft 2. Those editorial comments were as follows: * 38.3 - "shall" should not include signal status * 38.4 - contains broken reference * 38.4.1 - this section should be expanded as an informative spec. * Table 38.3 - can be removed (Table 38.4 covers this) * 38.5 - further elaboration on longwave transceivers is needed to support both singlemode and multimode fiber, this section should reference tables 38.6, 38.9, & 38.10 explicitly * 38.5.1 - needs expansion as an informative spec. * 38.6.1 - lift specifics of measurement techniques from FC-PH Appendix A. * Table 38.8 - can be removed (Table 38.9 covers this) * Table 38.11 - can be removed (redundant) * 38.6.4 - 1.5 dB is the wrong value, need to reference tables * 38.7.2 - redundant "1000 Base X" * Fig. 39.1 - 1000 Base-CX link, how do we differentiate the +/- connection since this is not a pin on the connector? Steve Swanson - Table 38.12 contains some inaccuracies. The limits for the dispersion slope, dispersion point, and link attenuation are all incorrect. Also, table 38.13 should have the reference to Fig. 38.3 in the 9 micron fiber column removed. Paul Kolesar - In Table 38.12, modal bandwidth should be labeled as "minimum requirement". The dispersion point should be labeled "zero dispersion wavelength". Additionally, in Fig. 38.4, the drawing is too manufacturer specific. The connection between the two SCs should be less descriptive. Fig. 38.1 needs to be resolved with the jitter document because it may be in conflict with the present jitter document. Dave Smith (Honeywell) - The BER of 10-12 is specified for the entire link. How do we propose to test this? Del Hanson - BER has to be specified in order to determine receiver sensitivity. This will be worked on by the task team. Jonathon's comments on the GMII - Electrical/timing requirements for the GMII are for the 10b chip scaled up to 1.25 Gb/s. The 1000 BASE T folks want to start over from scratch. There is a strong possibility that the GMII will not match the 10b interface. PRESENTATIONS ****************************************************************************** Del Hanson - Presentation describing the parameters that were used to perform the link modal calculations in order to determine the achievable link lengths. The rationale behind these parameters was described along with the set of conditions used to determine receiver sensitivity. ****************************************************************************** Ed Grivna - Presentation on impedance specifications. The tests identified in the existing draft will allow for operable links however they may not meet international test specs. The existing IEC connector specs do not specify this sufficiently either. A sufficient test spec. for measuring cable impedance has not been identified. Also, cable impedance tolerance has not been figured into the loss budgets. Differential TDR has been used to make these measurements but some people have questioned the validity of this approach. Is there a better method? Concerns were expressed regarding "daisy chaining" of cables. Present spec. does not prohibit this but it should not be allowed since it becomes difficult to control what the equalization will be. On the transmit interface, most implementations use capacitive coupling. Magnetic coupling is another alternative. Insertion loss is less than 0.1 dB through capicitors. NPO dielectrics should be used where possible to avoid ferroelectric effects. The Tx mask was presented which resulted in considerable debate regarding the meaning of the right and left axis. This requires further clarification. Equalizer tolerance effects were discussed as well as cable NEXT measurements and results. Using either a dual individually shielded pair or quad pairs, both were shown to have high isolation. HP test data from 11/96 confirms that there are no NEXT issues for either configuration. Cable shielding was discussed and it was explained that two shields are required to meet radiated emissions requirements. This configuration easily passes "Class B". ******************************************************************************* Steve Swanson - Presented a comparison of the various industry standards for optical fiber. The question is how do we best reference these specifications. Tables can be used to determine which document to reference but differences exist among the different specs which need to be reconciled. We also need to be aware of how the values will effect the link length calculations. ******************************************************************************* Karl Nakamura (LSI Logic) - Presented issues related to differential voltage levels for short haul copper. This presentation was the result of much discussion on the matter occurring over the reflector. A solution is being investigated that uses lower launch amplitude, lower receive sensitivity, and possibly lower loss cable solutions. This issue is expected to be resolved by the May meeting. ******************************************************************************* Christine Foster (W.L. Gore) - Presented performance data for DB9 interface. A 25 meter length of self equalized 21 gauge cable was used to obtain the data. With an input of 1.1V, 25-1 PRBS, @ 1.25 Gb/s, an 800 mVpp eye diagram was demonstrated. ******************************************************************************* Harry Takada (Sumitomo) - Presented a 3.3V longwave transceiver that operates at 1.25 Gb/s. Transmit & receive eye diagrams were shown at 1.25 Gb/s over the operating temperature range. This part is compatible with the Gigabit Ethernet proposal for longwave transmission, and demonstrates the feasibility of 3.3V operation which reduces power consumption. ******************************************************************************* Mark Nowell (HP) - Presented experimental data which supports the accuracy of the HP link model described at the January meeting. Particular attention was focused on the effects of mode partition noise in the link. Mark noted that the bandwidth measurement data was taken using the test laser, not an overfilled launch condition. Steve Swanson: Was the overfilled loss bandwidth measured? Mark: Yes but I can't recall what it was. I could provide it to anyone who is interested. Mr. X: Exit rise time is a function of fiber bandwidth for overfilled launch. To what degree is this true for restricted mode launch? Mark: The 1300nm laser does provide an overfilled launch. The model has been verified to track this. ******************************************************************************* Mark Nowell - Presented the effects of dirty connectors on multimode links. This presentation was in response to one given at the Jan. meeting which raised some concerns about increased modal noise in restricted mode launch systems which contained dirty connectors. Several experiments were performed to explore this effect but no appreciable degradation was noted. The conclusion reached was that this issue is not likely to play a part in real world systems. Steve Swanson: Was 850 nm restricted mode launch loss measured? Mark: No, but it can be determined from extrapolation of 3 data points already measured. Steve: At 1300nm, the loss is higher for RML than for OFL. This seems counter-intuitive. Mark: These particular connectors were chosen for this very reason in order to investigate this effect. Paul Kolesar: The presentation states that connectors can be permanently damaged by dirt when mating forces are applied. Were the ends examined to determine the nature of the damage? Mark: No, they were thrown away. Greg Smith: Does increased reflection from dirt increase RIN penalty? Mark: Yes Paul Kolesar: How restricted was the RML? Mark: Very underfilled launch. ******************************************************************************* Mark Nowell - Presented modal noise measurements made using Sumitomo 1300 nm transceiver. Paul Kolesar: Would you conclude that the present modal noise figures are appropriate? Mark: Yes, we are using what people are comfortable with. Concern expressed that we are over specifying the power penalties: Mark's response: Yes, possibly for 1300nm because it was increased to 2dB to account for dirt but dirt has been shown not to be a problem. ****************************************************************************** Rick Neumann (Vixel) - Presented data which was collected to confirm the present link budgets. Concluded that the 300 meter link length can be easily met with 850 nm VCSEL over 62.5 MMF and that receiver sensitivity should not be lower than -17dBm due to random jitter. For longwave link budget, 600 meters was achieved experimentally and jitter effects were well within spec. for the receiver. Also for the short haul copper budget, 25 meter operation was experimentally verified. Steve Swanson: What was the difference between OFL & RML measured bandwidth. Rick: Fiber BW decreases by 5 to 10% when RML is passed through mode scrambler. Also, results presented are slightly pessimistic because link length was derated to a particular length based on 160 MHz*km but this did not account for differences in attenuation. Rick's comment: The jitter data was taken to check the current assumptions on jitter allocations. Vixel's data disagrees with IBM's when increasing deterministic jitter as a function of length. David Cunningham: This might be explained by looking at the difference in spectral width between Vixel's and IBM's sources. Dave Smith: Was attenuator used to maintain same pulse amplitude at the scope for different power levels? Rick: Yes ******************************************************************************* Dave Smith (Honeywell) - Presented data from RIN measurements and also on the effects of launch spot size on fiber bandwidth. Concluded that a worse case RIN of -122dB/Hz is acceptable and that MMF bandwidth may be scaled to 1.25x if modes propagating near cutoff can be eliminated. Del Hanson - If we drop RIN to -122 dB/Hz then the RIN power penalty almost disappears. This is generally a good idea. In Fibre Channel, RIN is measured at DC using an RMS power meter. Should we be specing this measurement under modulation because this is more real world? Dave Smith: The danger of the current model is that we are taking worse case penalties for each parameter and adding them together without considering if they can add in this manner in a real link. We should keep an open mind about increasing the link lengths in the future. This could be achieved by increasing the Rx sensitivity by 0.5 or 1 dB. There are also concerns about the 400 ps transmitter rise time measured at 10 - 90%. Slow tails on the eye diagram could have negative impact on this measurement. Measuring this from 20-80 % would be more realistic. The Tx eye mask test is not as valid for MMF systems. It was originally designed for SMF (SONET) where fiber BW effects are small. The Tx should be measured against the mask with a receiver & filter which has a bandwidth representing worse case fiber with a 1 GHz receiver. I.E., need to measure what arrives at the receiver and so we simulate this with the combined filter and receiver. Schelto van Doorn: Fibre Channel specifies 20-80 % for lasers and 10-90 % for LEDs. General comment: Resolution of these issues is action items for the sub task group. ****************************************************************************** Peter Pepeljugoski (IBM) - Presented data based on IBM link model. In general, the power penalties are slightly worse than for the HP link model. The recommendation is to increase the available power budget and to use RML to increase the available link bandwidth. Another suggested alternative is to use a high bandwidth 62.5 MMF. Peter's comment: We propose to scale the BW by 1.25 to get 200 MHz*km. Rise and fall times should be set to 250 pS max. Del Hanson: 400 pS rise & fall time is more appropriate for the models based on distribution of measured parts. Paul Kolesar: BW scaling is a sleight of hand. Jonathon Thatcher: Not really because we are using the same 160 MHz*km fiber that is already widely in use. Steve Swanson: We have seen at least one fiber that exhibits no increase in BW due to RML. Jonathon Thatcher: We can't do use RML to guarantee that the fiber BW exceeds 200 MHz*km but we can add a scale factor per Dave Smith's proposal. MEETING ADJOURNED ************************************************************************* MINUTES: 1000BASE-T MARCH PLENARY Chair: George Eisler Editor/Secretary: Colin Mick ATENDANCE LIST (Taken afternoon of Tuesday, March 11) (last,first,company,voice,email) (PLEASE CHECK FOR ERRORS, MISSING NAMES) Agazzi,Oscar,Broadcom,714-450-8700,oea@broadcom.com Ali,Abe,SMC,714-707-2371,abe.ali@smc.com Almagor,David,National Semiconductor,408-721-8629,davida@galaxy.nsc.com Azadet,Kamran,Bell Labs,908-947-7280,ka@bell-labs.com Bender,David,Alteon,408-360-5510,dbender@alteon,com Benson,J. Paul,Lucent Technologies,770-798-2883,jpbenson@lucent.com Bestel,John,Lucent,610-712-7790,bestel@lucent.com Buchmeier,Brian,Pulse,619-674-8213,brianbuchmeier@pulseeng.com Cam,Richard,PMC-Sierra,604-415-6022,cam@pmc-sierra.bc.ca Campbell, Peter,Citicorp,(718)248-4199, Campbell,Robert,Lucent Technologies,908-957-2669,rrcampbell@lucent.com Choi,Francis,ADC.Fibermux,818-709-6000,fechoi@fibermux.com Chu,George,3Com,408-784-6031,gchu@nad.3com.com Coffey,Kelly,Lucent,408-980-3797,kellycoffey@lucent.com Darafshi,Bob,LSI Logic,408-954-3325,bdarafshi@lsil.com Debiec,Tom,Comm-Tek,717-335-7649,tom.debiec@ccm.usca.alcatel.com Diminico,Chris,Digital Equipment Corp,508-486-6983,diminico@lkg.mts.dec.com Dinh,Thuyen,Valor Elec,619-537-2633,tdinh@valorinc.com Easton,Doug,Comcore,818-880-5792,doug@comcore.com Eisler,George,Rockwell,310-459-9225,geisler@aol.com Essig,Dan,Rockwell,619-535-365,essig@brooktree.com Fifield,David,National Semiconductor,408-721-7937,fifield@lan.nsc.com Flatman,Alan,LAN Technologies,44-1260-297-966,a_flatman@compuserve.com Friedenbach,Ken,Sierra Research & Tech.,415-988-4806,ken@srti.com Froke,Rich,PICAZO Communications,408-232-9121,rfroke@picazonet.com Gladstone,Bruce,Comcore,818-880-5792,bruce@comcore.com Goldberg,Arie,Omnitron,714-250-6510,arie-ost@ix.netcom.com Gu,Richard,TI,972-995-7399,gu@hc.ti.com Guidon,John,Comcore,818-880-5192,john@comcore.com Halsey,Lloyd,Motorola,512-891-3294,lloydh@oakhill.sps.mot.com Hatamian,Mehdi,Broadcom,714-450-8700,oea@broadcom.com Hayden,Kirk,Level One,916-852-2871,khayden@level1.com Hinrichs,Henry,Pulse,619-674-8208,henryhinrichs@pulseeng.com Hochstedler,Charles,Lucent,610-712-5462,hochstedler@lucent.com Hsu,Jacob,Edge Point Networks,408-720-8500,jacob@edgepoint.com Joven,Juan,Level One,, Kairis,Vic,Nano Pulse,408-353-1753,vic4npi@aol.com Khuu,Tuan,Ratatornics,909-928-7700,tuan_khuu@datatronics.com Kim,Yongbum,LANart Corp.,617-444-1994,ykim@lanart.com Kozilek,Josef,Siemens AG,44-897-224-5311,josef.kozilek@vs.siemens.de Lacerte,Rick,Cabletron,603-337-5015,lacerte@ctron.com Langlands,Gordon,Wolfson,44-0-131-664-9386,gordon@wolfson.co.uk Laudon,Michael,Cypress Semi,408-432-7043,mkl@cypress.com Lee,Howard,BelFuse,510-227,0102 Leung,Tommy,Asante,408-455-8401,tleung@asante.com Lin, Chan-De,Macronix International,011-886-3-5788888,cdlin@mxic.com.tw Lorscheider,Jorg,Krone, Inc.,714-493-5952,75462.2666@compuserve.com Lussade,Jean-Pierre,Sony,408-955-6515,Jean-Pierre_Laussade@mail.sel.sony.com Makarem,Rabih,SMC,714-707-2422,makarem@west.smc.com Metzger,Steve,Prominet,508-870-5570,smetzger@prominet.com Miao,Tremont,Analog Devices,617-937-1222,tremont,miao@analog.com Mick,Colin,The Mick Group,415-856-3666,ckm@ix.netcom.com Milinarski,Fanny,Scope Comm.,508-393-1236,fanny@score.com Misaki,Tamihiro,Toko, Inc.,81-3-3727-1167,misaki@tojo.co.jp Murphy,Denis,Pulse,619-674-8317,denismurphy@pulseeng.com Naganuma,Ken,Toko American,847-635-3246,knaganuma@tokoam.com Ng,Vincent,SMC,714-707-2464,vng@west.smc.com Nobakht,Ramin,Rockwell,714-221-6862,ramin.nobakht@nb.rockwell.com Oh,Stepehen,National Semiconductor,408-721-2326,oh@galaxy.nsc.com Okada,Mitsuji,NEC Electronics,408-588-6549 Oktay,Ozay,Cannon,714-438-7180,ooktay@canon.com Oleynick,Gary,Berg,619-634-0723,goleynick@aol.com Pagnon,Andrew,National Semiconductor,408-721-6480,andres@galaxy.nsc.com Parker,Jim,Level One,916-855-5000,jparker@level1.com Raghavan,Sreen,Comcore,619-535-0074,sreen@comcore.com Rizk,Ramez,Nano Pulse,714-529-2600,rrizk@attmail.com Rose,Bonnie,Amphenol,416-291-4401,brose@cosmosgroup.com Sakai,Prescott,Cypress Semi,408-493-2756,pts@cypress.com Schebye,Tom,Fujitsu Microelectronics,408-922-9517,tschebye@tmi.fujitsu.com Sendelbach,Lee,IBM,507-253-1039,lsendelbach@vnet.ibm.com Shih,Cheng-chung,Allied Telesyn,408-523=5310,cshih@centre.com Stacy,David,HolonTech Corp.,408-369-4646,dns@holontech.com Telang,Vivek,Crystal Semi,512-912-3368,vivek@crystal-cirrus.com Thompson,Barry,Sierra Research & Technology,415-988-4802,barry@srti.com Thompson,Doug,OHMTech,714-752-1628,dougt@ohmtech.com Vafindes,Todd,National Semi,207-775-8905,todd@thoreau.nsc.com Verheggen,Bill,Intel,503-264-8594,william.c.verheggen@ccm.jf.intel.com Wang,Greg,Delta Products,510-668-5163,dpc-greg@deltaca.com Wang,Yun-Che,Cypress Semi,408-943-6817,ycw@cypress.com Wu,Robert,Accton,0110886-35770-270,rwu@accton.com.tw Wunner,Ken,U.S.Robotics,801-320-7690,kwunner@mhz.com Yeh,C.T.,Accton,0110886-35770-270-698,ctyeh@accton.com.tw Yousefi, Nariman, Broadcom,714-450-8700,yousefi@broadcom.com Yu,Mark,Lucent/Bell Labs,908-949-2185,my@lucent.com EVENTS OUTSIDE FORMAL MEETINGS: Tutorial rehearsal held Sunday evening 8-10pm. Tutorial held Monday, March 10 8-9:30pm. 250+ attendees. 350 handouts "consumed" during the week Presenters: Intro: George Eisler Status: Colin Mick DSP Intro: Jeyhan Kargouz Cabling issues: Chris DiMinico/Bob Campbell Enhanced TX/T2: Sailesh Rao CAP12: Mark Yu QAM 25: Mehdi Hatamian Dual-Duplex: David Almagor PR 9: Sreen Raghavan 802.3 gave final approval to 1000BASE-T PAR on Thursday, March 13. 802 Executive Committee approved the 100BASE-T PAR on Thursday, March 13. Forwarded to REVCOM (Colin has copies of the tutorial handout) FORMAL SESSIONS TUESDAY, MARCH 11 (PM) Presentations: Status report: Colin Mick Spreadsheet Status-Stefan Wurster QAM-Mehdi Hatamian Discussed FEXT issues Discussed jitter model issues Do we need BER simulations? Dual Duplex-Stephan Oh No data presented CAP12-Mark Yu Partial Response-Sreen Raghavan Provided data on primary proposal, some discussion of Dual Duplex possibility Enhanced TX/T2 proposal by Sailesh Rao with discussion of other proposals. Dan Dove raised the issue of external noise, suggesting that the group characterize external noise on legacy cable. Lucent team members expressed concerns expressed about Spreadsheet Table entries. After some discussion, Colin Mick proposed/summarized the following process (genral agreement by the group, no vote taken.). 1. Addition/deletion of rows in the spreadsheet requires a formal vote of the group. 2. All changes to cells should be made by Stefan Wurster. Request for changes should be sent to him. 3. Stephan will transmit the spreadsheet to Colin when it is ready for posting; Colin will post it to the reflector. Bob Grow (XLNT Designs, Tom Dineen (LSI Logic) and Haluk Aytac (HP) made short presentations to the group summarizing work/issues in Clause 35 (GMII). The bottom line is that 1000BASE-T will have to live with the electrical spec that is being developed NOW in in draft 2 of Clause 35. Specification data will be posted to the reflector. Implementors should review 35.3 (clocking) and the eelectical specification tables published in Clause 35 (d2) ASAP, definitely before the May interim. Kelly Coffey (Lucent) volunteered to serve as a liaison to the 802.3z GMII group for electrical interface and clocking issues. (Colin has hard copies of all presentations discussed above except those related to Clause 35.) WEDNESDAY, MARCH 12 (AM) The group began by reviewing individual SS table entries with a goal of identifying contested data points but found it frustrating. The group then turned to identifying "showstopper" issues/concerns for each of the proposals. Showstopper issues are listed below. GENERAL ISSUES: Jitter metric Hybrid loss Impulse noise tolerance Power limit Launch level, EMI & Noise tolerance Simulations should be conducted with timing recovery loop closed Prototypes/prototype-generated data are strongly encouraged Gate count metric/rationale OM jitter tolerance w/o impairment Noise BW DSP speed Block diagram/clock boundary/latency Clocking scheme Effectiveness in removing colored noise General assessment of 100BASE-TX/1000BASE-T complexity SPREADSHEET ISSUES: Crane test-TX/T2 (sorry, no more details) Magnetics frequency limits-how define BW A/D resolution ENHANCED TX/T2 (AKA PAM5+) ISSUES Margin for given parameters A/D resolution Combination Viterbi/DFE Jitter (general) Launch level @ noise immunity-EMI TAP count/gate count @ 3/10 dB Base Line Wander CAP12 ISSUES Block coding-packet length Latency values--FIFOs are missing 250 MHZ (7bit) in 0.35 micron CMOS? Jitter details 130 mv peak-to-peak BWS analog filer @ 100MHZ in 035 micron CMOS? QAM 25 ISSUES Need definition of 8-dimensional code-delimiters? latency? Mapped or table lookup Latency budget should include FIFO & delimiters Aggressive numbers BWS analog filer@ 100MHz in 0.35 micron? Jitter (10e-10) DUAL-DUPLEX ISSUES S/N margin There was general agreement within the group that at present the Dual-Duplex proposal is not a line code proposal. Individual line code teams can incorporate a dual-duplex signaling system into their line code proposals if desired. PARTIAL RESPONSE 9 ISSUES Block coding/Ethernet packets Latency budget-no transmit filter FIFO, R/S encoder DSP speed Base Line Wander WEDNESDAY, MARCH 12 (PM) NOTE: The plan for the afternoon was to define/assign action items deal with the "showstopper lists" developed in the morning session. Unfortunately, the afternoon work was splintered due to calls to meet with 802.3z to deal with "connector wars" ballots. (George and Colin will be pushing for completion of this task via email over the next month) The following agreements were reached before the group broke up. Jitter Metric-Sreen Raghavan volunteered to work on this Hybrid Loss-Pulse provided some hybrid data and retails on how the data were generated. Proposers agreed to incorporate the curve into their models for simulation (10KHz-100MHz). Copies of the data are available from Colin (will have to be mailed as will not FAX well.) Impuse noise tolerance-TX value is a good point of reference Launch level, EMI & Noise tolerance-Each proposer should make a statement relative to all three: (Crane test peak-to-peak voltage is __________ with a launch level of ____________peak-to-peak and for this launch level, the transmitter spectrum is ________________ (expressed as a spectrum diagram for 0-1 GHz.)) General assessment of 100BASE-TX/1000BASE-T complexity--There was considerable discussion of this issue. Eventaully the group voted to add a new column to the spreadsheet to indicate the additional gate count required to implement 100BASE-TX. VOTES: Move to add an additional item to the spreadsheet to indicate the additional gate count to implement 100BASE-TX. M-Jover, S-Mick Procedural motion Y-12, N-11, A--5 PASSED STRAW POLLS Meet with 802.3z for the May Interim (see details below) Y-15, N--6 INTERIM MEETING DETAILS Dates: May 15-16 (802.3z meets 14-16; as of March 12, 1000BASE-T has only agreed to meet May 15-16) Location: Ft. Lauderdale Marriot Details: $125/nite Make reservations by 4/22 800-343-2359 954-771-0400 ************************************************************************* 13. Connector presentations Wednesday am, 3/12/97 *************************** 13.1.1. Multi-Gigabit Copper Connector for 1.25/2.5 Gbs Ethernet - MetaGiG I/O ------------------------------------------------------------------------------ Ed Cady (Berg) 150 ohm impedence, up to 8 connectors, internal equalization in connector if required. Full shield, flexible bulkhead mounting, low profile footprint 0.87% NEXT, 0.18% FEXT, 30ps risetime(13 Ghz bandwidth) Berg is working with AMP on this (second source) 13.1.2. High Speed Serial Data Connector(HSSD) ---------------------------------------------- Lisa Huff (AMP) Same connector interface as Berg above. Data rates up to 2.1 Gbps Spec'ed in FC-PH-3, SSA-PH2, SFF-8460, IEC approval expected in May 1-3% NEXT unequalized, 3-4% equalized (George ? measured up to 10%) HJ noted that 24 gauge shown in plots, we spec 22? Lisa responded that connector supports 30 to 22 gauge cable. All measurements done with 2 pair (the others are power/ground, etc) 13.2.1. User Needs For Fiber Optic Connectors --------------------------------------------- Bob Fink (Lawrence Berkley National Labs) Basic requirements: industry standard, duplex, inexpensive, jumpers to others Doesn't have to be: cheapest, smallest, same as anything else, replacement Don't necessarily need another connector!!! DON'T need a connector war to slow down the standard 13.2.1.5. Fibre Channel Requirements for Fibre Optic Connectors --------------------------------------------------------------- Shelto Van Dorn Low cost (50% of duplex SC) [number 1 requirement] Components, installation (factory & field) No cost transfer between connctor & receptical Size (50% of duplex SC) [number 2 requirement] -high density, PCI mezzanine height, fibre spacing 0.5 mm minimum Duplex/polarized Multimode/singlemode Compliant with ISO/IEC 11801 and TIA 568A All source technologies No keying Dust protection Ergonomics Technical risk [number 4 requirement] Multiple sources Question: what is driving need for new connector? Don't think end users are bringing this. 13.2.2. Mini-MT Connector Interface ----------------------------------- Jim Kevern (AMP) Mission: increased density, reduced cost, transceiver interface Current Connector/Cable supporters: Siecor, AMP, Fujikura, Furukawa, Sumitomo Current Transceiver supporters: AMP, HP, Vixel, Fujikura America Current Ferrule supporters: US Conec, Fujikura, Furukawa/Europtics, Sumitomo NTT uses MT ferrule and MPO for Japanese telephone network Going to IEC, MT already approved Half the size, cost of duplex SC Telco apps to drive volume up/cost down No patents on ferrule, so no licensing required. Licensing for housing from NTT/US Connect 13.2.3. Mini-MT from Transceiver Vender's Perspective ----------------------------------------------------- Steve Joiner (HP) Respond to market, minimize number of optical interfaces supported Key to pricing is volume Concerns over reliability of Galaxy connector Question: isn't cost the major factor? how much do you pay for performance? Answer: this standard is for gigabit (not fibre to desktop), therefore, he feels that reliability, etc. are more important than lowest cost solution. Question: doesn't robustness mean more to the desktop? Didn't catch answer. 13.2.4. Galaxy Optical Interface/Connector ------------------------------------------ Tad Szostak [actually Sidney Berglund] (3M) Video demonstrating features of this connector Socket uses mechanical snap, gripping (no adhesive), V-groove to align fibres Plug & Socket style, no adapter Fibre bow provides optical contact pressure RJ-45 style latching Cost is 1/7 of a duplex SC Question: return loss is -20dB, which is right at the limit Answer: Yes, hope to do better Question: long term contaminant of fibres and in V-groove. Answer: they filled connectors with dust, then cleaned them, and they were OK Many other questions... 13.2.5. Mechanical performance of Silica Clad Fibre in Galaxy connector ----------------------------------------------------------------------- Thomas Hanson (Corning) Checked various stresses on fibre, lots of data End result: with 20 year life, stripped end has a 10ppm failure probability For gigabit application, the shorter height actually has a larger bend radius 13.2.6. LC Connector -------------------- Paul Kolesar (Lucent) Develop one connector for ALL applications in network RJ-45 style, slightly smaller Polarized Color Coded (SM blue, MM beige) Pull-proof Cable Compatible (field mounted on 900um buffered, 250 um coated) Performance: Meets TIA 568A, ISO 11801, *and* Bellcore GR326 Guarantee end-face contact High Reliability: proven tech, no bends, easily cleaned, low risk PCI mezzanine compliant (small version) Also have a new cord (not needed for connector): 1.6mm simplex, also available in duplex & quad configurations IEC 86B WG6, submission approved by USTAG Beta test with 3 sites, > 10k connections. Full production later this month 13.2.7. Fibre Jack connector ---------------------------- Richard Akins (Panduit) Only one with current work in TIA: FOCIS 6 (PN371) This design is exactly the size of RJ-45, and is flush-mounted The only design which can accept POF as well as GOF ATM accepted Fiber Jack for POF Uses same 2.5mm ferrules used by SC and ST today They are in full production now; >20k connections so far -------- end of connector presentations ------- 13.5. Sub Task Force Status *************************** Clause 4 (MAC): HF fixed a few minor pascal bugs possibly extend burst length; we need to see additional information in May Clause 41 (Repeater): SH Repeater State Diagram: remove wait state, use CRS as collision indication, add transition from preamble to idle Partition State Diagram: track .3u maintenance Clause 42 (System Topology): SH Add link length table for Full Duplex Update link distance limitations for 40 ByteTime latency in 1000BT Clause 30 (Management): DL Must match up with repeater state machine Only one new counter: aBursts, for repeaters only. New enumerations for phy types Clause 35 (GMII): BG Clause 4 consistency Incorporate clause 22 management registers Open comments primarily on electrical; significant progress, included in next draft Make register set consistent with clause 37 Make changes to clause 6: update figure showing which stacks each supports Clause 36 (PCS): RT D1 comments resolved (some left for D2) D2 comments worked on Ben Brown presented changes to PCS state machines, found in RTd2com_draft.pdf Call for simulation of state machines (by others); they appear pretty solid now, but would like non-designers looking at it. Comments to D2 to be updated within a week, ready for "early draft" for review Clause 37 (Link Configuration: FibreAutoNegotiation): RT Substantial changes to resolve nomenclature differences with clause 36 Technical changes: Link break function State machine timers/counts for handshakes Management registers Delete multiple PMA support Pause clarification Resolution of all these items will take several weeks, but ready for May Clauses 38,39 (PMDs): JT Created sub-teams 1000Base-CX: Ed Grivna (Cypress) Copper Cable: Christine Foster (Gore) 1000Base-SX/1000Base-LX: Del Hanson (HP) Optical Cable: Steve Swanson (Corning) PMD reflector, administered by Shelto Van Dorn PMA team with RT, work with DL on management requirements Hope to not drop length objective of 25 meters, instead increase receive sensitivity to 300 mV Progress on link budgets, need to resolve TBDs Clause 40(1000BaseT): GE 4 line codes; need to get to 1, spreadsheet developed Need input from GMII electrical and clock distribution Will request vote to approve PAR, will join rest of .3z in Florida interim State of Draft, Progress: HJ prepare for WG draft: review D2, submit comments probably release a D2.1 as change pages soon, then D3 as full set near May Draft & Working Group Process: David Law Description of balloting process, rules, voting, percentages, etc. 14. New Business (a.k.a. Motion Madness) **************************************** Motion #1 --------- Direct the editors to produce P802.3z/D2.1 in the form of change pages, for distribution prior to our May interim meeting, based on the resolution of comments accepted by the sub task forces at this meeting. Moved: Bob Grow Second: Tom Dineen Technical: >= 75% Friendly Ammendment by Ed Grivna: add the text "with complete D2.1 available on web in PDF" accepted. New motion is: Motion #1 revision A -------------------- Direct the editors to produce P802.3z/D2.1 in the form of change pages, for distribution prior to our May interim meeting, based on the resolution of comments accepted by the sub task forces at this meeting, with complete D2.1 available on web in PDF. RS: why the use of change pages? HF: it saves paper, and he wants D2 to live for a while. RT: clause 36 wants to release soon, clause 37 later; he would like to do it twice. HJ: email early changes yourself, D2.1 will be released close to the May interim. PT: can change pages appear on web as well. It will be done. SM: what is the target date? HF: sometime prior to May interim; probably 1-2 weeks before. PT: if subclauses get updated sooner, can they be put on web then? HF: normally, everyone works to a deadline, so nothing is likely to be ready much before the entire thing is ready. The Chair asked for a voice vote on this, expecting little opposition: Voting on Motion #1A (technical, needs >= 75%) Motion 1A Passed by voice vote, with 1 no vote ============================================================================== Motion #2 --------- Contingent upon IEC approval of their ongoing project for the ANSI Fibre Channel Style 2 copper connector interface, add to the current draft of the IEEE-802.3z section 39.5.1 page line 32, the following words: The style 2 balanced cable connector for balanced cables is the 8 pin shielded ANSI Fibre Channel style 2 connector with the mechanical mating interface defined by IEC ___. +-----------------------------------------------------------+ | Editor Note: (to be removed prior to final publication) | | The IEC number is scheduled to be available by the end of | | May 1997 and should be installed above ASAP. | +-----------------------------------------------------------+ Moved: Mark Sankey Second: Ed Grivna Technical: >= 75% Mark: This is the HSSC connector shown in both presentations today. ?: This is an addition, not a replacement of the DB-9 Andy Luque: What is approval date from IEC? Lisa Huff: Draft Standard in May 97, don't know when complete. Friendly Ammendment by David Fifield: insert "39.8" after page before line number, change "current draft" to "D2.1" accepted. New motion is: Motion #2 Revision A -------------------- Contingent upon IEC approval of their ongoing project for the ANSI Fibre Channel Style 2 copper connector interface, add to draft D2.1 of the IEEE-802.3z section 39.5.1 page 39.8 line 32, the following words: The style 2 balanced cable connector for balanced cables is the 8 pin shielded ANSI Fibre Channel style 2 connector with the mechanical mating interface defined by IEC ___. +-----------------------------------------------------------+ | Editor Note: (to be removed prior to final publication) | | The IEC number is scheduled to be available by the end of | | May 1997 and should be installed above ASAP. | +-----------------------------------------------------------+ HF: clarify intent to editor - if IEC assigns a project number (gets a PAR), put that number in where the blank is now. HJ: Uncertain if IEC will have a standard prior to 802.3z publication. Ed?: Project number in May, typical length is 9 months. Lisa Huff: previous motion in November stated "or in FC-PH" HF: This only applied to draft 1 of 802.3z. Concern is that if we finish our draft before IEC finishes, the number still changes -- this happened with the 802.3u MII connector. Lisa Huff: The DB-9 has an IEC number, but it only goes to 3Mhz, so it doesn't really cover 802.3z, even though we refer to it. JT: Concern that this is necessary but not sufficient. Does this mean that the rest of document is changed by clause editors? Mark Sankay: Intent is to get this connector into the standard. Andy Luque: The motion is too specific, should just state we insert connector into standard, not just in certain places. Friendly Ammendment by Howard Johnson: add "with other adjustments as needed to the draft to support use of this connector." accepted. New motion is: Motion #2 Revision B -------------------- Contingent upon IEC approval of their ongoing project for the ANSI Fibre Channel Style 2 copper connector interface, add to draft D2.1 of the IEEE-802.3z section 39.5.1 page 39.8 line 32, the following words: The style 2 balanced cable connector for balanced cables is the 8 pin shielded ANSI Fibre Channel style 2 connector with the mechanical mating interface defined by IEC ___. +-----------------------------------------------------------+ | Editor Note: (to be removed prior to final publication) | | The IEC number is scheduled to be available by the end of | | May 1997 and should be installed above ASAP. | +-----------------------------------------------------------+ with other adjustments as needed to the draft to support the use of this connector. Lisa Huff: Since IEC is point of contention, can we reference FC document? HF: Prefer IEC. Ed Grivna: Avoid referencing FC-PH, it is still in ISO process Ed?: There is also a EIA committee doing work on tolerences, etc. IEC will look to them. Voting on Motion #2B (technical, needs >= 75%) Yes: 86 No: 5 Abstain: 59 Motion 2B Passed with 94.5% approval =============================================================================== Motion #3 --------- To consider additional Fibre Optic connectors for inclusion in clause 38, at this meeting. Moved: Howard Johnson Second: Payne Technical: >= 75% Andy Luque: Does this mean this meeting only -- what about future meetings? HF: We voted to make connector decisions prior to working group ballot (July); if we don't make a decision on this today, (i.e. if we waited until May), we won't meet our schedule. PT: Could do a future project to incorporate a different connector for 802.3z HF: This is possible, but not probable (can't remember it happening in .3) SH: Since if this fails and connectors come up again, then we WILL fall off our schedule, wants to change motion to disallow any further motions before WG ballot. Steve Swanson: should we reword to say: "to continue to consider" HJ: Intent of motion is to see if there is 75% interest in putting in ANY new Fiber Optic connector in the standard, so we don't waste time on other motions if there isn't. Dan Dove: Is this really technical? HF: as chair, he gets to decide, and he says it is. Voting on Motion #3 (technical, needs >= 75%) Yes: 120 No: 34 Abstain: 20 Motion 3 Passed with 77.9% approval =============================================================================== HF: To keep this orderly, we will take a straw poll. The connector with the most hands gets to motion last, least hands first. JT: Does this mean that we could add as many as four new connectors to our standard? HF: Yes, since the motions probably will not be exclusive. HJ: mechanics: due to ordering, probably all after the first one which passes will pass as well. Ed Grivna: No open venue after presentations to ask questions of all vendors at once. Will one be available? HF: Ask when individual motion comes up Straw Poll of Fibre Optic Connectors: Connector count order Mini-MT 48 3 Galaxy 68 4 LC 8 1 FibreJack 15 2 HJ: There will probably be ammendments to motions, probably similar motions. Please refrain from mixing up connectors with ammendments (a'la Fibre Channel). This process was worked out with connector manufacturers, to their agreement. =============================================================================== Motion #4 --------- Modify clause 38.4 to additionally permit use of the LC connector. Moved: Howard Johnson Second: Bob Fink Technical: >= 75% Ed Grivna: Is the motion out of order? The copper connector had to be in IEC, why not the fibre ones? HF: The motion in November was for inclusion in the FIRST draft. We can include anything we want here, although people may have concerns about connectors not in IEC. Ed Grivna: Does have concerns about any connector which is not in IEC draft. Paul Kolesar: Lucent did not participate in process of straw poll/motion. He feels that these motions are just popularity contests, not based on technical merits. He is speaking against the process. HF: He did ask anyone who wanted to make a motion to discuss it over lunch. PT: Concerned about process, and the lack of discussion of issues other than during the presentations. HF: If there are concerns about the process, please offer a remedy. Dan Dove: Could we table this motion, then discuss the issues with the connectors? =============================================================================== Motion #5 (Privileged) ---------------------- Motion to postpone consideration until further discussion has taken place. Moved: Dan Dove Second: Bob Campbell Procedural: >= 50% JT: would like to ammend such that following this discussion, there would be no more discussion. HF: we can't discuss this according to roberts rules Friendly Ammendment (by JT?): replace "until further" and "has taken place" with "to allow further" accepted. New motion is: Motion #5 Revision A (Privileged) --------------------------------- Motion to postpone consideration to allow further discussion. Friendly Ammendment by HJ: add "for no more than 1 hour" accepted. New motion is: Motion #5 Revision B (Privileged) --------------------------------- Motion to postpone consideration for no more than 1 hour to allow further discussion. Voting on Motion #5B (procedural, needs >= 50%) Yes: 74 No: 55 Abstain: 24 Motion 5B Passed with 57.4% approval =============================================================================== Discussion of Optical Connectors: Paul Kolesar: What are our requirements? Seen FC requirements, have we adopted them? It seems ridiculous to choose connectors without them. Tad Szostak: Connector requirements were reflected in several of this morning presentations and mirrored those of Fibre Channel James Kevern: Fibre Channel went thru a 4 month process to come up with requirements. Since this group is 4x the FC group, it could take 4x as long. David Cunningham: We should base our requirements on the spec, tables in our clauses, including connector return loss, etc. ?: Propose that each proposal be up front to answer, in forum style. Ed Grivna: We are users of the technology, not developers. Concerns that data on the mini-MT was really taken on MT, and that it has recently changed. How much of the data is real to what we are voting on today? Jim Kevern: All data presented today was on mini-MT. Some data presented to the FC group was from MT. RS: requirements presented were size, cost, etc. none of which we have a set of requirements for, or measurements of. Hard to select between alternatives without criteria. Rick Akins: on mini-MT could you explain field polishing procedure and cost with 2 fibres? Todd Hudson: Based on a single ferrule, using special ribbon cable, process of strip, install and polish both fibres at one time. Field termination of this connector is at it's infancy. The process shown just proves it can be done. Rick Akins: What is the estimation of time/cost for field termination? Todd Hudson: Not known at this time James Luther: Don't take what was shown last night as final process. Rick Akins: What about tools to create Galaxy w.r.t. licensing? Tad Szostak: 3M will provide licensing in accordance to ANSI policy. Rick Akins: All that is required is the interface dimensions, so they don't have to license tools. ?: Transition to 40 vendors, 40,000 installers, better to stick with standard components, otherwise everyone will have to re-tool - which may never happen. Paul Kolesar: Agree with observation. It is pointless to choose a connnector in this committee. All are good, but shouldn't use this committee as leverage to show which connector is best. HF: Why weren't comments like this made earlier? PT: Major problem for FC vendor is passing EMC. How can we make a knowledgable decision without data on transceivers? We should wait. HJ: Not a good idea to select any connectors now. Focus on our interests, avoid risks. This is a very fast-track schedule. Any connector can be used. System vendors may ignore our recommendation anyway. Risks: connectors at early stage of development. When going thru IEC, they are likely to change. We may choose the wrong one. We are not experts in connectors. We should wait until it is obvious which one "won" - volumes, etc. Del Hanson: We should judge connectors against 20dB return loss. Total 1.5dB connector loss for all connectors in link. RS: Best standards are adopted, not designed. Technology is immature. Avoid standardizing on an immature product. =============================================================================== Motion #6 --------- Move to reconsider motion #3 (both mover/seconder previously voted in favor) Moved: Rich Seifert Seconded: Ross Procedural: >= 50% Point of order: Not all people who voted on motion #3 are in the room any more. HF: the meeting is still in session, so it's OK Steve: point of clarification - if this motion passes, we then revisit #3, which is a technical motion (even though this one is procedural). ?: Feel people are getting cold feet. Feels it is important for IEEE to adopt a new generation of connector. Right now fibre is not at par with copper. Success depends upon density. Recommend to make the difficult decision, not get cold feet. HJ: Glad to reconsider this motion. JT: Disagree. Voting on Motion #6 (procedural, needs >= 50%) Yes: 91 No: 57 Abstain: 12 Motion 6 Passed with 61.5% approval =============================================================================== Motion #3 reconsidered: Dan Dove: Not fair to lump all four proposals together when talking about maturity. Some are risky, some take current technology and step it down. Also, cost should be consdered. Steve Swanson: Thinks we should reword it to require 75% to overturn previous. HF: That's not what Roberts Rules state. Mike Salsman: Speaking against motion - no further connectors in standard. Most of us are system vendors. Most of differences in connectors are wall sockets, etc. Doesn't wish to delay the standard; do we need higher density at gigabit? Bob Campbell: Commend connector manufacturers. Voting today is premature; we don't have a set of requirements, especially reliability. This is very important. Mart Molle: If we don't choose a connector, people can wait to see what good connectors are. RS: Call the question. 2nd consideration voting on Motion #3 (technical, needs >= 75%) Yes: 55 No: 66 Abstain: 17 Motion 3 now Failed, with 45.4% approval =============================================================================== HF: now we have seen a connector war, and we don't like them! Cleanup: HJ: Motion #4 is withdrawn due to above decision. Dan Dove yields the rest of his time from motion #5 due to above decision. =============================================================================== Motion #7 --------- Not to entertain further new fibre optic connector/interface discussion in IEEE 802.3z Moved: Tad Szostak Second: Bowerman Technical: >= 75% HF: Intent is to say we won't discuss Fiber Optic proposals again in .3z Ben Brown: Agree, don't want new connectors, due to delay. But if this motion passes, can we hear updates on how connectors are progressing? So, he is against the motion as written. Dave?: Can't work with mixed messages. Don't want to work both in market and here in standards. JT: Thanks to connector companies for the entertainment and meals. David Cunningham: concerned that there are interface discussions which we need. Friendly ammendment: Remove the slash between connector and interface. accepted. The new motion is: Motion #7 Revision A -------------------- Not to entertain further new fibre optic connector interface discussion in IEEE 802.3z Tom: In favor of motion. There is no time in schedule for more connector wars. Steve Joiner: Also concerned that this motion may limit discussion of new optical interfaces. PT: Clarify what it says: .3z won't talk about connectors; if a year from now a new connector looks really good, start a PAR in 802.3 to add it. HJ: this motion only pertains to .3z, there may be future groups to add connectors. SM: Agree with sentiment, wording is unfortunate: Friendly ammendment by SM: replace the word "discussion" with "proposals". accepted. The new motion is: Motion #7 Revision A -------------------- Not to entertain further new fibre optic connector interface proposals in IEEE 802.3z Del Hanson: Since the four connector proposals are already on the table, can we discus them? Scott: suggested a rewording which was shot down. (and I missed it) Friendly ammendment by HF: strike "further new", add "any", and "other than the duplex SC" accepted The new motion is: Motion #7 Revision B -------------------- Not to entertain any fibre optic connector interface discussion in IEEE 802.3z other than the duplex SC. Steve: Friendly ammendment: strike proposals. Not accepted Dan Dove: call the question Voting on Motion #7B (technical, needs >= 75%) Yes: 119 No: 5 Abstain: 23 Motion 7B Passed with 94.4% approval =============================================================================== 15. Plans for next meeting ************************** Interim meeting May 14-16, 1997 (Wednesday-Friday) 8:30-5:00pm Hosted by Adaptec Fort Lauderdale Marriot North 6650 North Andrews Avenue Fort Lauderdale, FL 33309 (954) 771-0440 or (800) 343-2459 Project 802.3ab will co-locate (if room) July 7-11, 1997 - 802 Plenary week, Maui --Assume there will be a Monday morning meeting in July Sept (week of 8th), 1997: London Interim (3Com is working on this) November 10-14, 1997: 802 Plendary week, Montreal 16. Approval of January Minutes ******************************* Changes to January minutes: (HF will update them) January minutes were approved by voice vote, no opposition 17. Adjourn *********** Motion to adjourn. Passed by acclamation at 6:28pm (IEEE Social started at 6:30pm - whew!)