Date: September 7, 1990 To: Cable Working Group Participants From: Lawrence J. Lamers Subject: September 5, 1990 Cable Working Group Meeting Larry Lamers called the meeting to order at 1:00 p.m., Wednesday, September 5, 1990. He thanked Roger Cummings of StorageTek for hosting the meeting. The attendance list was circulated. Participants were requested to add or verify their facsimile numbers so they could be contacted if needed. They were also requested to indicate the number of nights they were at the hotel. As is customary, the people attending introduced themselves. The final agenda was as follows: 1. Single-ended Best Case Analysis (90-123 R1) [Spence] 2. High-density Connector Retention (90-106) [Chan] 3. Color coding cables (90-136,137) [Blackford] (90-138) [Wagner] 4. Parameters, Procedures and Philosophies (90-134) [Fiala] 5. 68-conductor Cable Construction 6. Next Meeting The following people attended the meeting: Name Status Organization ------------------------------ ------ ------------------------------ Mr. Denis D. Springer O 3M Company Mr. James Fiala S 3M Company Mr. Charles Brill P AMP, Inc. Mr. Peter M. Blackford P Astro Cable Company Mr. Thomas Debiec O Belden Wire & Cable Mr. Wills Xu O C&M Corp. Mr. Raul Ugalde Moncivais O Centro de Invest. Condumex Mr. Jose Manuel Villasenor O Condutel Mr. Chuck Micalizzi A Emulex Corp. Mr. I. Dal Allan P ENDL Mr. Terry Maezawa P Furukawa Electric Amer, Inc. Mr. Kenneth Post P Future Domain Mr. Kurt Chan P Hewlett Packard Co. Mr. Mike Peper A Hewlett Packard Co. Mr. Howard Wang O Hitachi Mr. George Penokie P IBM Corp. Mr. Gary R. Stephens A IBM Corp. Mr. David A. Buesing O IBM Corp. Mr. Chuck Grant A Madison Cable Corp. Mr. Lawrence J. Lamers P Maxtor Corp. Mr. John Lohmeyer P NCR Corp. Mr. David Steele S NCR Corp. Mr. Gerald Houlder A Seagate Technology Mr. Paul E. Orland V Solbourne Computer, Inc. Mr. Robert L. Simpson P Sony Corp. of America Mr. D. W. Spence P Texas Instruments Mr. Peter Dougherty P UNISYS 27 People Present The following new documents were distributed at the meeting: Document Doc Date Author Description of Document ------------- -------- --------------- --------------------------------------- X3T9.2/90-123 8/31/90 B. Spence Single-Ended Cable Best Case Analysis Rev 1 X3T9.2/90-127 8/16/90 K. Chan 16-bit Connector Retention and 32-bit Physical Layer Thoughts X3T9.2/90-134 8/24/90 J. Fiala Rough draft of cable test parameters, procedures, and philosophies X3T9.2/90-135 9/4/90 B. Spence "November" Working Group Announcement X3T9.2/90-137 8/29/90 P. Blackford Proposed SCSI-2 color coding appendix X3T9.2/90-138 8/31/90 D. Wagner SCSI-3 Color Code The results of the meeting 1. Single-ended Best Case Analysis (90-123R1) [Spence] Bill Spence presented a document on interim results of his continuing study of single-ended applications. He made the following recommendations based on these results: 1) don't put the initiator in the middle of the bus 2) pulling the bus down in the middle is the worst case 3) using active termination greatly improves system integrity 4) during a data transfer one of the devices should be at the end of the cable. His document presents an analysis of voltage vs. time as a signal is released on the SCSI bus. In particular he looked at the possibility of double-clocking REQ and ACK. The legal best-case scenario indicates that a minimum impedance of 79 ohm is necessary to stay out of trouble. With 110- ohm termination 79-ohm impedance is also acceptable; with 132-ohm termination 90-ohm impedance is required. John Lohmeyer asked if totem-pole drivers would be a better choice to improve the negated voltage level? The consensus was that they are, but more than the drivers affect system integrity. The negative spike is what causes problems, and it comes from impedance mismatch and stubs. Slowing down the edge limits the negative glitch and ringing. This is done by turning off the transistor softly. 2. High-density Connector Retention (90-127) [Chan] HP is using the high-density connectors and as Kurt Chan stated in his document they have some concerns regarding the retention mechanism. Kurt proposed that we consider adopting the IPI/HiPPI connector for the SCSI-3 P-connector. His reasons for this are: 1) the break-away requirement is un-substantiated in real life 2) there has been a significant precedent set with other interfaces 3) the reasons for SCSI-2 8-bit divergence from IPI no longer valid in the 16-bit high-performance environment 4) testing at HP has uncovered compatibility and quality problems with the current retention mechanism because it is not specified in the standard 5) not all retention mechanisms intermate well and some plastic versions break-away once and break. The jack-screw retention mechanism proposed is a simple extension to existing high-performance connector family. HP modified a SCSI-2 connector by adding a standoff that has 4-40 thread into the bulk head and uses a 2-56 thread for the thumb screw retainer. They are working with AMP to finalize the design at this time. The design allows use of either clip retainers or jack-screw retainers according to Chuck Brill of AMP. Kurt recommended that the jack-screw retention mechanism be added to SCSI-3. He will provide dimensions for standoffs and implementors notes on bulkhead mounting. 3. Color coding cables (90-137) [Blackford] (90-138) [Wagner] At the request of the committee Dick Wagner and Peter Blackford developed a proposal for color-coding SCSI cables, both 50- and 68-conductor. Included in that proposal were the lay-up construction used for each. Peter presented foils of the color coding scheme. He noted that the proposed scheme meant that single-ended and differential cable assemblies were different. John Lohmeyer had a negative reaction to this statement, pointing out that NCR today does not require different cable assemblies for single-ended and differential cables. He did not to differentiate the two markets merely to get a standard color on each signal. Peter responded that it is possible to use the same cable assembly for both as is done by some folks today. However, it was his understanding that it was desirable to keep the REQ/ACK signals in the inner core for both environments, and this was not possible because the differential pinouts are not the same as the single-ended arrangement. Current single-ended cable assemblies can often be used in a differential environment because differential has greater noise margins and thus is less susceptible to the cross-talk. Chuck Grant of Madison stated that they use a 3-9-13 lay-up, and claimed it works better than a 2-8-15 lay-up. Wills Xu of C&M agreed with this claim. Of the other two vendors present Furukawa uses 2-8-15, and Belden uses 3-9- 13. That was an even split. However with the proposed pinout to cable pair assignment this does not appear to be a real issue. In a 3-9-13 lay-up the ATN signal goes in the inner layer along with REQ and ACK. The data signals are still in the outer layer along with TERMPWR. Dal Allan postulated two questions: 1) will manufacturers accept a standard color coding for cable? 2) why should the connector pin assignment to signal wire color codes be specified? A question was raised on whether REQ and ACK should be together in the cable. This might result in a lot of crosstalk between them. Bill Spence noted that crosstalk when the signal pulses very close together is not a problem and if they are far enough apart it also is not a problem. John Lohmeyer stated that in a synchronous transfer the signals can have any timing relation since they are independent of each other and not interlocked. There was no resolution on whether this might be a serious problem or not. Bill Spence requested that the committee develop a separate cable specification and cable assembly specification. Chuck Grant questioned why the color coding of IPI and HiPPI were not being used. Peter Blackford responded that the color coding was optimized for the 50/68 pin SCSI cable. No existing color coding was adopted to keep a level playing field. The color coding scheme adopted was based on Western Electric's scheme which is preferred by assembly houses. Dal Allan stated that he has seen Madison color codes at several vendors using IPI and HiPPI. Dal requested Peter to submit a revision of his document that did not tie the signal pair and color coding to a connector pin-out. Chuck Grant accepted an action item to provide Madison's color code scheme, designating colors and pairs but not signal pin-outs for the next X3T9.2 mailing. An item will be placed on the plenary agenda to address SCSI cable color coding. 4. Parameters, Procedures and Philosophies (90-134) [Fiala] Jim Fiala handed out his document on testing cables. This is the long- awaited study of test methodology. It is a formidable document. It also included his recommendations on parameters and values for single-ended SCSI cable applications. While the parameters were generally the same as agreed to at the last Cable Working Group meeting the values for those parameters were more stringent. Jim said the definition of test methods permits the tighter specification. Jim stated that at present only Madison and Berk-Tek have provided cable samples for testing. Bill Spence expressed an interest in nailing down exactly which parameters really determine a good cable. (This has been the objective of the Cable Working Group since its inception.) He is continuing his investigation into this at Texas Instruments. Dal Allan picked up on this point and wanted to know what are the three best tests and the values of parameters for those tests that determine a good cable. Jim Fiala responded that the document is organized in order of effectiveness of the testing methods for determining a good cable. Jim also made a point that the cable preparation is critical to getting good and consistent results. His document specifies a two-page long procedure for preparing cables. Peter Blackford stated that it is more likely to screw up a good cable with marginal assembly techniques than vice-versa. Kurt Chan supported this statement, noting again the testing he had done at HP. Dal Allan pressed the issue of cable assemblies and the need to test those. The end user customer does not buy cables per se, but cable assemblies that include connectors. Jim Fiala stated that a correctly applied connector should not affect the testing results. Dal noted that test fixtures have been developed in other industries that are go/no-go testers for cable assemblies. No one knew of such a tester for SCSI cables. 5. 68-conductor Cable Construction This topic is covered in Peter Blackford's presentation (see document 90- 136). The lay-up is 5-11-18, with all the data signals in the outside layer. However, the parity bits are in the middle layer whereas in the 50- conductor cable, the parity bit is in the outer layer. There were no negative comments, so this construction seems to be acceptable to both users and vendors. 6. Next Meeting The next Cable Working Group meeting will be held on Monday, October 29, 1990, starting at 1:00 p.m., in the Stouffer Austin Hotel, Austin, TX. The hotel telephone number is 512-343-2626. The meeting is hosted by D.W. (Bill) Spence. On the agenda are: 1) A discussion of testing methods for inclusion in an appendix of SCSI-3 2) Resolution of the color-coding scheme for SCSI cables (if the plenary meeting doe not resolve it).