Date: September 9, 1989 To: X3T9.2 Membership From: Lawrence J. Lamers, X3T9.2 Recording Secretary John B. Lohmeyer, X3T9.2 Chairman Subject: September 6-7, 1989 X3T9.2 Working Group Meeting John Lohmeyer opened the meeting by thanking Gene Milligan of Imprimis for hosting the meeting in Oklahoma City, OK. John announced that X3T9 had forwarded SCSI-2 Rev 10b to X3 with a sufficient vote to make it stick this time. He expected that the document would begin public review in October. One comment has already been received regarding the high-density connector selection. It will be dealt with at the appropriate time. The complete 10b document will not be distributed in the mailing. A document detailing the changes from revision 10 will be mailed out, so that a person can mark the changes in his/her revision 10 document, if desired. The final agenda was as follows: 1. Additional SEND DIAGNOSTIC Pages (87-186R0) [Spence] 2. More Than 8 Devices On Wide SCSI (87-207) [Harms] 3. Search Command Modifications (88-002) [Stai] 4. Autoconfiguration 5. Autosense 6. Request For Data DAT Device Type [Andrews] 7. Documentation Layering [Stephens] 8. Single-Cable 16-bit Wide SCSI [Penokie, Lohmeyer, Lamers] 9. 16-bit Everything Proposal (89-97) [Houlder] 10. Alternate Physical Layers (e.g., fiber optics) 11. LOGICAL UNIT RESET Message Proposal (89-058) [Lohmeyer] 12. Arbitration Fairness (89-61) [Penokie] 13. Arbitration Unfairness (89-65) [Buesing] 14. Single-Ended Terminator Power Analysis (89-115) [Chan] 15. Target Initiated SDTR Message Control (89-110) [Gerry Houlder] 16. SCSI Working Group Schedule 17. SCSI Electrical Special Working Group Meeting Report The following people attended the meeting: Name Stat Organization ------------------------------ ---- ------------------------------ Mr. Robert C. Herron A 3M Company Mr. Denis D. Springer O 3M Company Mr. Joe Lawlor P AT&T Mr. John D. Walden O EI DuPont Inc. Mr. Chuck Micalizzi A Emulex Corp. Ms. Jean Kodama S Emulex Corp. Mr. I. Dal Allan P ENDL Mr. Robert Liu P Fujitsu America, Inc. Mr. Kenneth Post P Future Domain Mr. Kurt Chan P Hewlett Packard Co. Mr. Eric Tausheck V Hewlett Packard Co. Mr. George Penokie P IBM Corp. Mr. Gary R. Stephens A IBM Corp. Mr. David A. Buesing O IBM Corp. Mr. Gene Milligan P IMPRIMIS/CDC Mr. Lawrence J. Lamers P Maxtor Corp. Mr. John Spongr A Mitsubishi Electronics Amer Mr. Gary Murdock V National Semiconductor Mr. John Lohmeyer P NCR Corp. Mr. Brian Scanlon V Seagate Technology Mr. Fred Burgess V Seagate Technology Ms. Rita Lin P Sony Corp. of America Mr. Robert N. Snively P Sun Microsystems, Inc. Mr. D. W. Spence P Texas Instruments Mr. Jeff Stai P Western Digital Mr. Doug Pickford A Western Digital Mr. Erik Jessen V Western Digital 27 People Present The following new documents were distributed at the meeting: Document Doc Date Author Description of Document ------------- -------- --------------- --------------------------------------- X3T9.2/89-36 9/6/89 G. Stephens Preparations for new physical layer(s) Rev 1 X3T9.2/89-94 9/5/89 G. Penokie 16 Bit Data Path on a Single 68-pin Rev 2 Connector Results of meeting 1. Additional SEND DIAGNOSTIC Pages (87-186R0) [Spence] Under this topic, a discussion has been going on about the X3B7.1 project which is defining a Diagnostic Command Set (DCS) for SCSI. Document X3B7.1/89-19 contains a proposed diagnostic command set for testing SCSI devices. It is primarily for testing disk drives. A copy of this document will be included in the September X3T9.2 mailing as X3T9.2/89-119. The next X3B7 DCS working group meeting is October 27, 1989 in San Jose and the next X3B7 plenary meeting is November 6-7, 1989 in Phoenix. Contact George Canevit of ACT Technologies for details. Bill Spence commented that to reinforce the testing point of view Texas Instruments has been active in promoting this approach due to the experiences from staying at the leading edge of using disk products. He felt that TI needs the visibility into the drive (ala ESDI) to be comfortable with latest technology products. He was concerned that there are not enough people participating in the project to get broad industry support. Gene Milligan stated that after a quick review the document it appears to be too restrictive, not recognizing alternate drive technologies. Bob Snively pointed out that the functions of evaluation of the design, checking the drive at incoming inspection, and checking the functionality in the field can be quite different. Downloading different code is one way to achieve these different needs. A review of this working paper is planned for the second day (October 31) of the October X3T9.2 working group meeting. 2. More Than 8 Devices On Wide SCSI (87-207) [Harms] At Joe Lawlor's request, this proposal is to be merged with the single-cable 16-bit SCSI proposal. 3. Search Command Modifications (88-002) [Stai] Bob Snively accepted an action item to look into the feasibility of importing the boolean constructs from the IPI-3 search command. Jeff Stai insisted that his good name be removed from this item on future agendas. 4. Autoconfiguration This item will be dropped from future agendas until another proposal is received on the topic. Dal Allan mentioned that this topic will be part of the Fiber Channel discussions since such environments may require autoconfiguration. 5. Autosense Bob Snively still has the action item to develop a proposal on autosense, but he is waiting for SCSI-3 maturity. Bob requested that this discussion be included under the alternate physical layers topic (see item 10). 6. Request For Data DAT Device Type [Andrews] The item will be dropped from subsequent agendas until a proposal is received. 7. Documentation Layering (89-36R0) [Stephens] This was a controversial proposal. It was even difficult to achieve consensus over whether elements of this proposal are evolutionary or revolutionary. The underlying thrust of this document is to move to a packetized protocol. This requires that asynchronous processing functions be removed -- devices should process the packets after they are completely received. Dealing with reset conditions (both hardware and messages) caused a substantial discussion with the only agreement being that a lot of work is necessary. During the discussion, a revision 1 of this document was defined to remove some confusion over terms like "asynchronous". These changes should aid in understanding that Gary is referring to asynchronous processing and not to asynchronous data transfer. John Lohmeyer agreed to make these changes to the file that Gary provided. No resolution was reached other than to continue to discuss the proposal at the October plenary meeting. 8. Single-Cable 16-bit Wide SCSI (89-042R2) [Penokie, Lohmeyer, Lamers] George Penokie presented his latest proposal on single-cable 16-bit SCSI. The single cable is now called the "P cable". Section 4 additions were reduced. Gerry Houlder's proposal for COPY command modifications was included in section 7. George presented three alternatives for pin assignments: P1. The original pin assignments optimized for noise rejection and centering the terminator power lines. This causes some difficulties in mixing 8-bit devices as conductors must cross each other to get to the A cable pin assignments. P2. Use the A cable pin assignments adding the new signals all at the end of the cable. This un-centers the terminator power lines creating a possibility of damaging devices or opening fuses if cables are built backwards. It is also easiest to mix 8-bit devices with 16-bit devices. P3. Center the A cable pin assignments on the 68 conductor connector. Add half of the new signals at each end. This proposal retains centered terminator power and does not require crossing signals to map to 8-bit devices. After some discussion alternative P3 was selected. The RESERVE(10) and RELEASE(10) commands should be mandatory only for P- cable devices. 9. 16-bit Everything Proposal (89-97) [Houlder] This proposal was rejected because all compatibility with 8-bit devices and with existing protocol chips would be lost. 10. Alternate Physical Layers (e.g., fiber optics) This topic was discussed extensively under agenda item 7. Autosense is included in Gary's proposal. 11. LOGICAL UNIT RESET Message Proposal (89-058) [Lohmeyer] Several people expressed concerns about this proposal: Bob Snively: The situations where this message would be used do not occur very often and could be handled by other means. The overhead of supporting another resetting/aborting message should avoided. Overriding a reservation should be done via a RELEASE command. Larry Lamers: The initiator could change its ID to that of the failed initiator and issue an ABORT message. John Lohmeyer countered that if the failed initiator later came back on the bus it would be very confused about what happened. Gary Stephens: The proposal creates a security problem. If initiator A attempts to reset the I/O's and reservations from initiator B, the target should reject the attempted reset unless initiator B has permitted the target to accept resets from initiator A. A new enabling message would need to be defined for this purpose. Gary also preferred a solution that permits the outstanding I/O processes to be re-directed rather than aborted. Gene Milligan: The proposal should be expanded to include a way to include a dual-channel priority override. John Lohmeyer said he would review the concerns within NCR and possibly submit a revised proposal. 12. Arbitration Fairness (89-61) [Penokie] Deferred to next working group. 13. Arbitration Unfairness (89-65) [Buesing] Deferred to next working group. 14. Single-Ended Terminator Power Analysis (89-115) [Chan] We need to determine the VOL specs for the new termination. There is a trade-off between increasing the voltage/current to help the high-level noise margin, but then we may exceed the 48 mA current specification on the drivers. The difference is going from 48 mA to 52 mA. The pads should be able handle 68 mA. One of the questions that needs to be answered is, "what is VOL at 52 mA?" Kurt's document listed several alternatives and recommendations. Another question raised was whether the terminator power specification should be increased to 1000 mA. If so, we should also recommend a higher fuse value. Since very few people had reviewed this proposal, the issue was placed on the agenda for the October plenary meeting. 15. Target Initiated SDTR Message Control (89-110) [Gerry Houlder] In spite of Gene Milligan's valiant defense on Gerry's behalf, this proposal was rejected. 16. SCSI Working Group Schedule Ken Post is hosting the next meeting in Santa Ana on October 30-31, 1989. A CAM meeting is scheduled for November 1, 1989 in the same hotel. A Fiber Channel meeting will be held November 2-3 at Amdahl in Sunnyvale (still in California, but considerably further North). January 8-12, 1990 - tentatively hosted by Bob Snively in San Jose. March 5-9, 1990 - tentatively hosted by Jeff Stai in Irvine. May 7-11, 1990 - tentatively hosted by Joe Lawlor in Chicago. The 1990 meetings are scheduled for five days as follows: CAM Meeting Monday SCSI Working Group Meeting Tuesday & Wednesday Fiber Channel Working Group Meeting Thursday & Friday 17. SCSI Electrical Special Working Group Meeting Report Bob Snively reported on the meeting. Later, Larry Lamers provided the following report: "A meeting of the SCSI electrical special working group was held on the evening of September 6, 1989 in Oklahoma City, OK. Gary Murdock of National Semiconductor presided. Gary Murdock presented the agenda for the meeting: 1. Reach agreement on working group objectives. 2. Discuss fitting 16-bit SCSI onto standard 50-pin cable. 3. Discuss power and die size problems associated with integrating 10 Mbit/S transceivers onto 8- or 16-bit SCSI protocol controllers. 4. System requirements. 5. Other topics. 6. Schedule next meeting. 1. Integrating transceivers of differential capabilities into 8-bit and 16-bit SCSI is a desirable objective. An alternative is a separate differential transceiver chip or chips that minimize the space required on a printed circuit board. The general consensus was that differential transceivers were required to meet the transfer rate, distance and noise immunity characteristics desired. 2. The question of fitting 16-bit SCSI onto the current 50-pin cable was raised. The consensus was that this is not of great interest at this time. It may be possible to put 16-bit SCSI on the current 50-pin connector, using open collector on control signals, and differential on data and REQ/ACK. In the past Gerry Houlder had raised coupling issue with this type of approach when flat cable is used. Bob Herron of 3M stated that foil covered flat cable does not have this problem. 3. A discussion was held on the power and die size problems associated with integrating 10 mega-transfer per second tranceivers on 8-bit and 16-bit SCSI protocol chips. The ideal chip is one with integrated differential transceivers, little cost penalty, 16-bit SCSI, a plastic package, surface mount, and that does not require external heat sinks. The consensus is that the ideal chip is possible but hot; an ECL type chip could be done with todays technology although it would be expensive and run hot; a CMOS type chip would be marginal with todays technology and would probably require a ceramic package. Power requirements @5.5 volts 8-bit is 4 watts, 16-bit is 7 watts; @3.5 volts 16-bit is 4 watts, 8-bit is 2.5 watts. This indicates that reducing the Vcc would help the heat dissipation problem. Eric Jessen of WD said that it appears that the ISO and SCSI requirements for differential driver common mode rejection are not consistent. Reducing the common mode rejection would also help the heat dissipation problem. Kurt Chan stated that with some restrictions fast SCSI can be run in single ended according to his testing. However the group questioned why even use a single-ended approach if a differential 16-bit is possible. Gary Murdock stated that due to the limits of die size, power dissipation and complexity of an 18 channel transceiver it would be difficult to design and costly to manufacture. He expected that the yields would not be acceptable. The viewpoint that the group adopted is that of person going from single-ended to differential not going from existing differential implementation to the new differential. This makes cost considerations a higher priority. There was some discussion of staging the transition. An upgrade path of first going to 16-bit SCSI, then to 10 mega-transfers per second, and onto 20 mega-transfers per second with improved compatible silicon was suggested by Bob Snively. 4. The discussion on system requirements was short. Perfection was desirable but a 25 meter cable length, with 10 mega-transfers per second and typical error rates would be acceptable. The result of the meeting is several action items. The next meeting is scheduled for the afternoon of the first day of the November working group meeting. Gary Murdock of NS will re-visit the the IPI studies made in the past and investigate a two chip approach at 10 mega-transfers. Eric Jessen of WD will look at ceramic packages and investigate possibility of ECL approach. All chip vendors are requested to do an analysis based on releasing the following constraints: a) lower common mode noise rejection (1.5 volts below ground to 1.5 volts over Vcc), b) lower Vcc to 3.5 volts, c) eliminate counter drive requirements (an approach like RS-422 uses). d) package the transceivers separate from the controller. A request is made of all for information and studies to determine how much common mode noise rejection is needed?" There was considerable discussion on Thursday about the special working group. It is clear that the general group feels this work is important. Dal Allan expressed concern that the group seemed to be ignoring the approach of using a higher current single-ended driver in favor of the new kind of differential driver. He was concerned that the differential solution may take much longer than the single-ended solution. Other people maintained that SCSI is currently in a phase where we should be looking to the long term solutions.