MINUTES FOR IEEE P802.3z Closing Plenary SESSION WEDNESDAY, NOV 12, 1997 MONTREAL CANADA session starts at 1PM (no secretary for the first few minutes. Chair confesses to and appologizes for breach of order. Mark Sankey volunteers to act as secretary) subgroup reports: MAC (Howard Frazier). Only one minor technical change in clause 5 management Pascal to ensure that carrierSense test is performed only in halfDuplex. This is a cleanup task necessitated by 802.3x, which neglected to make this change since clause 5 was deprecated at the time 802.3x was done. Management (David Law): Minor change to wording of aSymbolErrorDuringCarrier (see below). GMII Grow (BG): There is no MII mode for GMII, but it is still possible to implement an interface that supports both. Added t_period maz of 8.5 ns and moved the frequency tolerance spec (f_clock = 125MHz +/- 100 ppm.) to the same table. 802.3ab wants to add a 1ns peak-peak jitter to this table as well. But they didn't write a comment. Howard Frazier (HF): I object. This is a process problem. We need to make it clear to the 802.3ab group that they need to read the sponsor draft. Geoff Thompson (GT): A linear ramp on these changes as we move forward is appropriate. HF: With regard to the 1000BaseT request of a jitter spec, we will place an editorial note in D3.3 which will reflect the requested spec, along with a statement that a comment will be submitted on the sponsor ballot to convert the contents of the note into a normative specification. Quackenbush: It may be the case that a garden variety crystal oscillator will not meet that spec. BG: There were also two substantive editorial comments, one on voltage tolerances and one on AC thresholds. Haddock: There were only two repeater comments, both editorial, 41 and 42. Rich Taborek (RT): There were changes in the arcs are represented in the state machines. There is now a separate arc drawn if the conditions are not the same. This has no effect on the functionality, however. There are 8 TRs altogether, 6 from Scott Mason. We need to get signatures on these. HF: Give us a summary of the technical changes made in resolving these TRs. RT: The first one is a three part problem (Comment 91) The rx state machine asserts carrier now if an I is converted to a C via a data error (distance is three). This fixes the first two parts. The early_end state now has an exit condition to cover receive I or C. This covers the third part. Comment 95 corrects a recent error that was introduce into the draft regarding carrier detection. Carrier detect is now asserted when either of the follow conditions occur: 2 or bit distance between the received data and either K28.5 or 2 to 9 bit difference between the received data and the expected K28.5 There was also a TR (Comment 54) concerning inconsistent terms for client of PCS. GMII?, MAC? This TR was rejected (?) To address TR Comment 54, change the wording to "for half duplex PHYs only) Then changed to "PHY's half duplex clients". Rich Siefert (RS) (who wrote TR) agrees on this wording. Comment 98 : Delete paragraph on PCS transparency. Comment 94: There was an error in the transition between Autonegotiation (AN) to client packets. This should be done in a frame synchronous manner to avoid making a partial packet look good. Scott still has not signed C99. David Law (DL): Sponsor ballot pool closes 11/24. The normal process is too slow to work for anyone who has not already started, but if you fill out a special form that I have, this will expedite the process of you getting added to the sponsor ballot pool. These forms can be gotten from me and must be returned to me by tomorrow night. There was a comment against Clause 30.3.2.1.6 line 44 aSymbolErrorDuringCarrier. When is it incremented? It was corrected in the case of full duplex only from slot_time to minimum packet size. HF: Bob Campbell's comment on clause 39 is next, since he has to fly out of here. Bob Campbell (RC): Currently there are two connectors defined for CX. Let's just chose Sytle 2. Impedance and NEXT are both better with Style 2 connectors. They are also smaller. Having one connector will eliminate much customer confusion. Motion 1 (2PM): Add to C39.5.1 "It is recommended that the Style 2 connector be used as the MDI connector. Moved: RC, Second: Ed Cady. Technical (75%) Discussion: RS: Why have DB9 at all? Howard Johnson (HJ): Many manufacturers are already making Style 1 cable assemblies. Let's not pass this. HF: Some of the people we really need in order to change this are not in the room. Tom Dineen: Didn't we already to remove the Style 1 connector? HF: No, we voted to add the other connector. We voted not to remove Style 1 in London. GT: My friendly amendment is to add a note which says: "The intention is to deprecate Style 1 in a future version of the standard." HF: Your ammendment doesn't sound friendly, because if it is included, the motion will probably be defeated. GT: Make it a hostile amendment then. I make the motion (2:20 PM). RS: Second Pat Thaler (PT): This is a mistake. This is too incremental. We are creating to many levels of disapproval. Let's wait for the EMI test results to come in. Andy Luque (AL): Call the question (on the hostile amendment) HF: Any objection to calling the question? None noted, the question is called. Vote: yes 4 no 27 abstain 14 technical 75% required. AMMENDMENT FAILS Ed Cady: Currently we make both connectors. The trend is toward Style 2. We are making many mixed jumpers. Please stick with one connector. HF: Any further discussion? None requested. Vote on the Campbell/Cady motion: VOTE: yes 55 no 6 abstain 19 Technical 75% required MOTION PASSES Del Hanson (DH): PMD subgroup report. There are two categories of comments: those that relate to Comment 754 and those that do not. The general approach has been to resolve all TRs. The non-DMD comments: Comment 79 (TR) - a reference to where the link budgets came from. This has not been signed off yet. Comment 185 (TR) - this is the one reqiring reference to international standards. GT has signed this off. Now the DMD-related comments: The link lengths have been maintained at: SX LX 62.5 MM 260m 440m 50 MM 550m 550m SMF na 3 km The jitter budget has been maintained at its currenty specified numbers. We have defined a conditioned launch and referenced an existing test Coupled Power Ratio test. Of the 4 C754 related TRs, one has been signed off (HJ's). Chris DiMinico:I will sign. GT: I want a discussion of reliable system operation. HF: Note that Joe Gwinn is not a voter, but we should follow up with him to determine whether he is satisfied with this resolution. Paul Kolesar (PK): I want to modify my TR, but I do not think that it is currently resolved. EG: Here is my Clause 39 subgroup report. There are 5 open TRs. The TDR measurement comment has a response pending. There is the return loss comment. Responses to all comments have been drafted, but some have not been signed off. HF: National Laboratoy Fiber Report. Surveyed fiber performance at 3 National Labs in California. Presentation materials will be made available on FTP site. Quackenbush: Do jumpers ever degrade fibers that otherwise work OK without them? HF: Without drawing too many conclusions, offset jumpers (for LX) seem to work. It is worth doing more investigation. Not one SX link within the specified length was found that failed to work. David Cunningham: Tutorial on Modal Bandwidth Investigation Attack problem by conditioned launch and offering guidance on receive bandwidth max. We need to avoid radial overfilled launch. We need to avoid exciting widely separated mode groups or just a few mode groups. Coupled Power Ratio (CPR) = power_mm_fiber - power_single_mode_fiber (in db) Panel of Optics Experts Q+A Del Hanson Jim Tatum David Cunningham Jonathan Thatcher Peter ? (from IBM) Don Knasel more Q: Can we build the launch conditioner into LX transceivers? Panel: We could, and may, but it would have to be removable, since handling smf and mmf in the same LX transceiver is normative. Panel: The situation we have it that the patch cord is always necessary for LX and almost never necessary for SX. Mark Sankey: If I make all the customers use patch cords for unconditioned launch SX transceivers now, will things still work when they later get a conditioned launch transceiver and launch into the patch cord? Jonathan Thatcher (JT): Yes, but it should not be standard operating procedure to put in jumper cables for SX. Q: Should be use jumper cable for single mode LX links? Panel: No. Michael Salzman: Do we use the jumper on the transmit side only? Panel: The jumper will have TX and RX in a pair for duplex SC. So the jumper would be present on the RX side as well, but not required. BG: How long before transceiver manufacturer's can make these modifications (for SX) Panel: They are already available from more than one manufacturer. Tom Dineen: These patch cords should be readily differentiated from others by labelling. Michael Salzman: What about if one of these jumpers only is used, say from rack to rack (no other fiber)? Paul Kolesar: You cannot put the LX cables in backwards. JT: All the patch cords we have gotten so far have been very well marked. DH: some labelling guidelines will be in the standard. JT: LX jumper will only work if the cable is not put in backwards. Q: Is there a maximum length for the LX patch cord? Q: Normally a wide range of jumpers are used. Can we get offset jumpers of these lengths? Panel: In theory, yes. DH: We will make a jumper app note. Peter ? (IBM): If you plan on long jumpers, as the main jumper, you would have to replace the long jumper if you changed wavelengths or changed to single mode fiber. HF: What is the plan for specifying good cable or respecifying old cable to a higher level? Panel: Improving the specs is certainly easier now that there is a constrained launch. HF: What do we have to do in 802.3 to get other standards bodies to work on an enhanced fiber spec? Panel: Chris DiMinico is going to push this with TIA via a letter to FT2.2. Joseph: Why don't we just recommend moving to SMF for the future. The fiber prices are about the same. Jim Tatum: There are a lot of cost reductions that are available at 850 nm multimode that you would not capture. Flexibility would be lowered since the LED stuff wouldn't work. DH: We still target LX:SX cost ratio at 2:1 for the same volumes. JT: 1394B gigabit is short wave 50 micron fiber. This will drive the costs even lower. HF: It is not 802.3's charter to recommend new fiber installations. DH: describes the 1,2,4 Gb/s 1394B path Q: Is an offset required for the patch cord to work properly? HF: Yes. Without offset, the patch cord did not work all the time in the National Labs tests. Dave Cunningham (DC): We're not defining a patch cord. There may be other methods to condition the launch. Scott Mason: What are the effects of bend radius? The ATM Forum picked up on this. DC: They picked up on worse case everything. Tatum: The ATM Newsletter said jiggling the cable would cause the system to fail. I have never seen this happen. JT: while this might be a concern for unconditioned launch, it is of no concern for conditioned launch. Scott Mason (SM): If we did change the jitter budget, what kinds of length improvement could be achieved? DH: We don not have enough data to reallocate the jitter budget. No group is ready to give up any of their margin. However, more on this can be done in the future. JT: Modal bandwidth is the real issue, not jitter. HF: Let's limit the discussion to things we can affect. Invites Ed Grivna to join panel of experts. EG: If this interface had a jitter problem, we would have already seen it in the tests. Decreasing the bend radius would actually tend to reduce DMD effects by redistributing modes! Very short jumpers are not an issue. DC: In all of these tests, the jitter was probably all typical. In our simulations we use worse case jitter. Wen-tsung Tang: Any potential problems beyond DMD? What will be next? DH: We thought we had a very solid spec pre DMD. We feel we have a very solid spec now. HJ: What is your overall level of confidence? Was this a quick fix or fundamental? DC: Concerning launch, we treat the problem in a fundamental way. We have been working on LX offset patch cords for two years. Moving to conditioned launch is a fundamental shift that will allow cable companies to make higher performance specs. HF: This may be the thing that GE is remembered by. It is original research. In the next few months, this will be verified and reverified. Q: Over the lifecycle of GE, source technologies may change. Is there anything about conditioned launch that could be a gotcha for a new source? JT: Hard to say, but this ought to help those designing new sources. END PANEL DICUSSION MOTIONS Paul Kolesar (PK): Patch cords will be expensive, so I want to make a motion - (PK) Motion: To allow separate LX multimode and LX single mode transceivers in the standard. (3PM) Second: Shelto Van Doorn PK: This way you do not always need a patch cord for LX multimode. Shelto: LX_multimode tranceivers could be made less expensive. Joel: Previous motions already say that fulfillment of the conditioning requirement may use an external conditioner. RS: This makes things more complicated. Realistically, most intrabuilding applications will be SX and most interbuilding applications will be LX single mode. PK: I agree it is more complicated, but it levels the SX/LX playing field. John ?: Can I ever adapt LX_multimode transceiver to single mode? HJ: I am against this. We decided this issue in Enschede. LX multimode "fell out" of our definition of LX single mode. Additionally, this breaks Clause 30 and 34. David Law: this breaks Clause 30. PK: We will fix these clauses. Scott Mason: What exactly does you motion do? It just allows LX transceivers that fail the single mode test. Mart Molle: Its like RS232. We could just label the transceiver. Do we have to change the standard? PK: Yes, we have to change it. DH: HP would definitely support this with a product. But it's up to the user. PT: There is a difference between Distinct Identity and deliberately competing products. DC: You could conceivable mate LX_multimode to single mode fiber with a special patch cord. Tom Dineen: I oppose this due to schedule concerns. HF: Any changes we make now threaten the consensus we have. Motion Number 2: 1) Modify Table 38.7 to make two subtypes: LX_M and LX_S change text in 38.4 introduction to describe LX_M and LX_S globally change LX to LX_type include in 38.8 introduction "an LX_S transceiver may be used on MM fiber if an external mode conditioner is used. The mode conditioner for LX_S transceiver shall conform to the LX_M specs of table 38.7" Moved by P. Kolesar Second: Shelto Van Doorn RS: Call the question HF: Any opposition to calling the question? PK: Opposition to calling the question Vote on calling the question: yes 45 no 16 Question is called. Vote on main motion: yes 30 no 48 abstain 2 Technical 75% required MOTION FAILS HF: I would like to entertain a motion from Howie Johnson: Motion Number Three: (4 PM) Accept the resolution of all comments as recorded in the database and as modified by motions passed 11/12/97. Make appropriate changes to the draft to produce D3.3 with editorial cleanup as required. Request that 802.3 adopt the attached motion. Attached Motion: Date: 13-November-1997 Having adopted resolutions for all comments received on D3.1 and D3.2 of P802.3z, the Task Force requests that 802.3: Affirm the work produced by 802.3z as reflected in D3.2 with the resolution of comments. Recirculate D3.3, (along with outstanding TR comments) Authorize 802.3z editorial staff to respond to editorial comments generated by the recirculation, and to produce D4 based on those responses. Request that 802 forward D4 for LMSC sponsor ballot, contingent upon the successful completion of the WG recirculation of D3.3. Authorize 802.3z to respond to comments generated during the sponsor ballot, and produce and recirculate D4.1 based on the resolution of comments. Request conditional submittal of D4.1 to RevCom for their consideration at the March, 1998 meeting, contingent upon the successful completion of the LMSC sponsor ballot, and recirculation, if required. M: H. Johnson S: B. Grow Discussion: BG: You are asking for conditional approval. Do we meet the criteria for this? HF: I believe so. There isn't sufficient time for an SEC electronic ballot between the close of the recirculation and the date when we must ship to the IEEE for Sponsor Ballot. HJ: Add "editorial to item 3 above" (Secretary's note: I already did.) HF: Any further discussion? None offered. Vote: yes 63 no 0 abstain 0 Technical 75% required Johnson/Grow MOTION PASSES Next Meeting Arrangements: February 2 and 3, Double Tree Hotel, Bellevue, WA. Hosted by Signal Consulting, Inc. Approve Minutes from Santa Clara meeting: Approved by Acclamation. Meeting Adjourned 5:00 pm.