X3T9.2/92-159 To: Membership of X3T9.2 From: D.W. Bill Spence and Lawrence J. Lamers Date: 30 July 1992 The SPI working group is the umbrella for all contact, cable, transceiver issues related to SCSI-3 Parallel Interface (SPI). Bill Spence chairs this working group which is chartered with developing a set of recommendations for the SPI standard that will get the physical plant to a reliable state. Agenda -- X3T9.2 July SPI Working Group Rochester, MN -- 21 July 1992, 2-8 pm 1. SILICON 1.1 Differential driver skew specs -- Chan 1.2 Skew budget review -- Chan 1.3 Differential driver output specs -- Gingerich 1.4 Differential common mode range -- Ham 1.5 Single-ended active negation, which lines? -- Ham 1.6 Capacitance on SCSI signal wires -- McGrath 2. CABLES 2.1 Test Procedures--Tom Debiec 2.2 Applicability of Rise-Time Degredation -- 3M or Spence 2.3 Justification for twisted-pair ribbon cable -- Ham, 3M 3. TERMINATORS 3.1 Nominal vs worst-case current sourcing limits; shall vs should -- Spence 3.2 Limit on terminator current demand from TERMPWR -- Spence 3.3 Technical presentation from TI 4. SYSTEM 4.1 Provisions re Hot Plugging -- Ham 4.2 Review of Annex C -- Spence 4.3 Review of Rev 6? -- Lamers 4.4 Presentation on Differential Performance -- Ham SPI Meeting Attendees -- July 21, 1992 Name Status Organization Phone Number ------------------- ------ ----------------------------- -------------- Mr. Norm Harris P Adaptec, Inc. (408) 945-8600 x2230 Mr. Charles Brill P AMP, Inc. (717) 561-6198 Mr. Bob Whiteman A AMP, Inc. (717) 780-7481 Mr. Jan V. Dedek P Ancot Corp. (415) 322-5322 Mr. Florin Oprescu O Apple Computer (408) 974-2354 Mr. Tom Debiec P Berk-Tek, Inc. (717) 354-6200 Mr. Bob Gannon O C&M Corp. (203) 774-4812 Mr. Michael Smith S Dallas Semiconductor (214) 450-0457 Dr. William Ham A Digital Equipment Corp. (508) 841-2629 Mr. Paul R. Nitza A Emulex Corp. (216) 236-8504 Mr. Tom Jones V Emulex Corp. (714) 668-5361 Mr. D. W. Spence A ENDL Associates (512) 255-0339 Mr. Howard Wang O Hitachi Computer Products (408) 986-9770 x207 Mr. Wayne Kosters V I-TECH Corp. (612) 941-5905 Mr. George Penokie P IBM Corp. (507) 253-5208 Mr. Otis Greene O IBM Corp. (914) 892-5597 Mr. Kevin Pokorney O Intellistor, Inc. (303) 682-6649 Mr. Chuck Grant A Madison Cable Corp. (508) 752-7320 Mr. Lawrence Lamers P Maxtor Corp. (408) 432-3889 Mr. Robbie Shergill P National Semiconductor (408) 721-7959 Mr. Joe Vo V National Semiconductor (408) 721-2614 Mr. John Lohmeyer P NCR Corp. (719) 596-5795 x362 Mr. Stephen F. Heil P Panasonic Technologies, Inc (201) 348-7064 Mr. Dean Wallace O Texas Instruments (214) 517-2623 Mr. Richard Mourn O Texas Instruments (214) 997-3426 Mr. Kevin Gingerich V Texas Instruments (214) 997-3378 Mr. Mark Jordan S Unitrode Integrated Circuits (603) 429-8628 Mr. Shishir Shah A Western Digital (714) 932-7235 Mr. Tak Asami O Western Digital (714) 932-7621 29 People Present Status Key: P - Principal Member A - Alternate Member L - Liaison Member O - Observer S,V - Visitor Results of Meeting Bill Spence called the meeting to order at 2:00 pm. He thanked Hoerst Truestedt of IBM for hosting the meeting. As is customary, the people attending introduced themselves. A copy of the attendance list was circulated for attendance and corrections. It was stated that this is an authorized X3T9.2 SPI Working Group meeting. The meeting was conducted under Roberts Rules of Order. Since this is a working group meeting no final actions are taken. The working group prepares a set of recommendations for approval by the X3T9.2 Plenary Committee on Lower Level Interfaces. The minutes of this meeting will be posted to the SCSI BBS and the SCSI Reflector within 14 calendar days. Attendance at working group meetings does not count toward meeting the attendance requirements for X3T9.2 membership. The general working group meetings are open to any person or company to attend and express their opinion on the subjects being discussed. The voting rules for the meeting are those of the parent committee, X3T9.2. These rules are: one vote per company; and any participating company member may vote. 1. SILICON 1.1 Differential driver skew specs -- Chan [92-114] Kurt Chan reported that the 6 ns skew number that is typical of external drivers is not the worst case because it doesn't account for different junction temperatures. The 10 ns worst case number is not acceptable from the system view because it does not allow sufficient skew for the protocol chips, etc.. Bill Ham cited a data integrity issue due to double bit errors caused by skew delays in differential operations. An uneasy compromise was achieved as reported in the next item. 1.2 Skew budget review -- Chan The revised skew budget is: Transmitter Chip 32 ns setup / 45 ns hold Foil 1 ns External Driver 8 ns ------ TX Connector ------ 25 meter cable 4 ns Distortion 4 ns Tolerance 0 ns ------ RX Connector ------ External Driver 9 ns Foil 1 ns Reciever Chip 5 ns setup / 15 ns hold This is still subject to further changes pending review by all parties. In particular, National Semi wants time to study the delay differences between transmitters and recievers. 1.3 Differential driver output specs -- Gingerich [92-123] Kevin Gingerich suggested that the Vol and Voh be removed because they do not belong in a diffrerential specification. He also proposed a test circuit. The working group agreed to these changes. 1.4 Differential common mode range -- Ham Bill Ham asked again if this range can be reduced to +/- 2 v keyed on 2.5 v. After hearing of the can of worms this issue raised the last time it was brought up, Bill quietly put the lid back on. 1.5 Single-ended active negation -- Lohmeyer [92-148R1] John Lohmeyer brought in a proposal to specify the characteristics that determine an active negation driver and to clean up some of the wording related to single-ended drivers/receivers. Paul Nitza and Tom Jones expressed concern that requiring 7 mA at 2.0 v worst case (or 2.0 v min at a test current of 7 ma, as the concept was first discussed at the April SPI meeting) may cause excessive power pin currents for devices that operate at the other extreme. Florin Oprescu stated that he calculated a need for 10 ma minimum for worst case to get to the 2 volt level on first step, but the corresponding voltage range did not become clear. Paul and Tom wanted action suspended on John's proposal, saying they had only had two days to study it. But John and Bill countered that the numbers being objected to are already in the SPI draft standard working document, and that John's proposal simply introduces some badly needed clarification. The working group agreed to recommend the proposal about as is and asked the Emulex folks to investigate their concerns for a future meeting. An editor's note should be included in the draft SPI document to indicate that the active negation numbers are still subject to debate. Bill Ham suggested that a maximum rise time spec is also needed. No action was taken pending a written proposal. 1.6 Capacitance on SCSI signal wires--Jim (Quantum) McGrath [92-143] Because it dealt with node capacitance matters touched on by John Lohmeyer's paper above, this paper was discussed next. The case for softening the node capacitance limit to accomodate switchable terminators [92-143] (McGrath) was discussed. There was a strong concensus that allowing increased capacitance in each SCSI device for the switchable terminators severely impacts the signal quality and reliability of a bus. The working group recommends that the 92-143 proposal (and, by extension, the use of certain switchable terminators, as now known) be rejected. 2. CABLES 2.1 Test Procedures--Tom Debiec [92-150] Tom Debiec proposed six procedures for measuring cable characteristics. Tom will add cable preparation requirements, and this revised document will be included in the draft. 2.2 Applicability of Rise-Time Degredation -- 3M or Spence 3M not being present, Bill Spence reviewed Jim Fiala's recommendation about rise-time degredation as a useful cable parameter [X3T9.2/90-134]. Little support for adding it to the test procedures developed, however. Florin stated that the rise-time degradation can be calculated if the attenuation is measured up to 200 Mhz. Rise-time degradation will not be added as a cable spec. The attenuation measurement will stay at 5 Mhz. 2.3 Justification for twisted-pair ribbon cable -- Ham Bill Ham stated that his lab tests of twisted-pair shielded round cable vs unshielded flat cable showed no substantial difference. Bob Snively cited some history that indicated the twisted-pair flat cable addressed crosstalk issues. George Penokie also cited a real life example of a problem fixed by use of twisted pair. No action. 2.4 One Cable for S/E and DIFF P and Q cable -- Ham [92-138] Bill Ham presented proposed wording to facilitate use of the same cable assembly for S/E and DIFF P and Q cables. No objections. 3. TERMINATORS 3.1 Nominal vs worst-case current-sourcing limits; shall vs should -- Spence Bill campaigned for the softening of the current sourcing limit. If it really affects anything, the MTBF change should be cited. Bill noted the IBM/Aeronics experience sinking 60 ma. Several of the silicon folks were still very nervous about this. The draft document should have a note that a plenary vote will decide. 3.2 Limit on terminator current demand from TERMPWR -- Spence The TERMPWR fuse specifications in the Standard are based on the max current demand made by the original resistive terminators or the Boulay terminator. Any increase in terminator current demand will cut into the operating margin of the fuse. The standard already has provision to control current demand from TERMPWR by non-terminator circuits, but a need for a limit on demand from the terminars was not anticipated. In the face of development of a number of more complex terminator circuits, however, this topic needs examination. There was considerable support for introducing a limit based on the demand of resistive and Boulay terminators. Any vendors affected should be alert to the resulting impact. 3.3 Technical presentation from TI -- Dean Wallace [92-160] Dean Wallace made another presentation on TI's new constant-current mode terminator, which sources 24 ma into any line whose voltage is below about 3.0 v, working from any TERMPWR voltage from 3.5 to 5.5 v. It is switchable and adds 2.5 pf per line capacitance. In general, the scope pictures looked real good. Mark Jordan of Unitrode cited their experience with current mode terminators. Problems occurred which were attributed to constructive interference in the middle of the bus. The resulting glitch is a result of the energy in the cable. The current flows in a cable longer than when a Boulay terminator is used because the energy is not absorbed. One needs to analyze the system from an energy viewpoint to understand what is happening. Discussion brought out a concrete example: when a host at a bus end originated an -ACK pulse, a disk at the other end of the bus saw a good quality -ACK pulse, but a disk in the middle of the bus did not. TI has not seen that result with their design, but it is now on the record as something to test for. 3.4 Termination Sensing Proposal -- Florin Oprescu [92-161] Florin's concern is with a bus which ends at a bulkhead connector, where an end user may connect an optional bus extension and otherwise should install an external terminator. He described a design which may be installed on the bus just inside the bulkhead and which senses whether or not there is a terminator installed downstream from itself. If not, it automatically turns on a terminating function. According to Florin, this proposal does not require a change in the SPI draft. It will be considered for an appendix. 4. SYSTEM 4.1 Provisions re Hot Plugging -- Bill Ham [92-137] Bill brought in a formal proposal for an implementation note of some kind on how to hot-plug successfully. It was not discussed for lack of time, but it was in line with Bill's earlier reports and their discussion. 4.2 Review of Annex C -- Spence There was little input as the hour was getting late. 4.3 Review of Rev 6 -- Lamers Bill Spence and Larry Lamers plan to construct revision 6 for consideration at the next plenary meeting. 4.4 Presentation on Differential Performance -- Ham Bill Ham reported on test results which seem to affirm that differential SCSI is indeed more ruggd than single-ended, despite its tighter skew budget. Future meetings -- There is no separate SPI meeting scheduled for Seattle, but John proposes to devote a considerable part of Tuesday afternoon, August 18th, to SPI matters. The SPI Chairman is scheduled to be out of the country at the time of the September meeting; if a meeting occurs anyway, it will be announced on the SCSI bulletin board, the SCSI Reflector, and the SPI FAX roster. It would be wonderful if there were no need for any more SPI meetings by October, but I wouldn't count on it. The meeting was adjourned at 8:00 pm.