IEEE P1355 Working Group Third working group meeting 1st September 1993 Minutes The third meeting of the IEEE P1355 Working Group was held at Bull SpA, Pregnana, Milan, 10.00 - 17.00 on 1st September 1993 Present Ralf Bok„mper Harting ElektroniK GmbH Stefan Haas CERN Ernst Kristiansen SINTEF Andrew Cofler Bull SA (Editor) Calogero Mantellina Bull SpA Paulo Melloni Bull SpA Pino Rosario Harting Elektronik SpA Peter Thompson INMOS Limited Colin Whitby-Strevens INMOS Limited (Chairman) Apologies Roland Marbot Bull SA (Co-Chairman) Sylvain Paineau THOMSON RCM Georgio Bonvini Molex 1 Welcome, introductions, approval of agenda CWS welcomed everyone to the meeting, thanking Bull for their hospitality. The agenda was proposed and approved. It was requested that the roster be put onto the FTP server. ACTION CWS: put WordPerfect version and ASCII text version onto the FTP server. 2 Approval of minutes of 22nd June meeting The minutes, previously distributed, were approved as a correct record. 3 Chairman's report 3.1 Sponsorship CWS reported that there was no change of status since the previous meeting 3.2 Press Release The Press Release had been made simultaneously by several partners, and with the IEEE Computer Society (who published an abbreviated version, consistent with their guidelines). Good articles had been published in several journals, and several good contacts received. It was agreed that a further press release should be considered at the end of the year, when plans for taking the draft to ballot will be drawn up. It was proposed that a file of press articles should be maintained. ACTION ALL: send to CWS copies of Press Articles, particularly those appearing in national press. ACTION CWS: dsitribute the collated press articles with the proposed distribution of the next draft. 3.3 Boston/Danvers (BASC) meeting preparation CWS reported that he was awaiting confirmation that the tutorial has been scheduled for Monday 18th October, and a one- day WG meeting on Tuesday 19th October with BASC in Danvers, Boston. It was agreed that the tutorial should be presented by CWS (rationale, DS links), AC and/or Roland Marbot (gigabaud links), RB or Richard Carlson (connector/cable) and EK (application to SCI). At the end of the meeting, it was agreed that AC, PT and CWS should prepare a revised draft by 1st October, which should be mailed to all on the roster for review at the WG meeting in Boston. 3.4 Ballot preparation CWS reported that he has had initial contact with the IEEE on the formation of the ballot group, and he explained the balloting process. It is important that all members of the WG have some form of IEEE membership, so as to be able to take part in the ballot group. AC reported that at the recent Hot Interconnects symposium, IEEE were making membership special offers. AC questioned the timetable to ballot, particularly as there are some issues concerned with 3 Gigabaud signalling, and with optics, where there is much work to do to develop the proposals and to verify that the proposals are reasonable. It may not be desirable to hold up the balloting process until all of these issues are resolved. CWS confirmed that the current plan is to review these issues at the end of the year. If necessary, the WG could consider addressing these issues in a subsequent standard (eg a .1 standard), which, if the scope has to be changed, would require a re-submission to NESCOM, both for a revised scope to P1355, and for a new .1 WG to be created (which the IEEE would regard as a separate project with a new PAR). 3.5 Other contacts CWS reported that substantive technical comment had been placed on the reflector by Paul Walker, and that he had received contacts from AT&T concerning optical connectors, from CNN regarding cable construction, from Ferranti, concerning using DS-Links and HS-Links, from Norden Technologies, concerning various DS-Link issues, and from Sylvain Paineau concerning optical connectors. Relevant technical material was copied and distributed at the meeting. 3.6 Organisation reports and specifications AC reported that there was interest in the Bullit evaluation chip for HS links, and requested whether it would be possible to place the data sheet on the FTP server. CWS undertook to check whether this is agreeable from the IEEE point of view. He noted that the reports from CERN evaluating the DS-Link technology had been presented as P1355 working papers, and had been very well received. There is a wider issue concerning references to trade-marked names within the standard document, which CWS would also explore. ACTION CWS. 4 Review of draft standard CWS tabled a revision of the overall structure and the DS-Link related components of the draft. AC tabled a revision of the protocol hierarchy and the HS-link components of the draft. CWS reported that, regrettably, it had not proved practicable to combine these into a single (machine readable) document before the meeting. Many minor comments, and typographical errors, were noted for editorial correction. The following significant issues were discussed:- Contents Sections 5.3 and 5.5 on DS-Link optical interconnect should be separated out, as they do not use DS encoding, and a number of issues have to be handled separately. 2 References The reference to draft IEC 48B(SEC)229, should be replaced by IEC 1076-4-101 consistently throughout the document. Reference should be included to Fibre Channel.. 3 Definitions The definitions for link input and link output should be generalised away from VLSI devices (they also apply to connectors). A link interface has functionality, as well as comprising a link input and a link output. 4 Protocol Stack The diagram needs changing: i) DS-Link optic differs from DS-Link copper at the exchange level. In fact, this is best considered as a different technology, and the name DS-Link not applied. The provisional name TS-Links was invented (for `Three of Six' encoding). ii) there will (may) be different speeds defined for HS links, which will differ at at least the signal level. A consistent terminology needs to be invented to describe the various options at the physical level. The following was suggested:- -- where is one of DS, TS or HS, is SE (for single ended electrical), DE (for differential electrical), FO (for Fibre Optic), and the - is the speed in 100's of MHz (01 for 100 MHz, 10 for Gigabaud, 30 for three Gigabaud). DS-Links do not need a speed specification (see current proposal). Thus, currently, the following are specified in the draft:- DS-SE DS-DE TS-FO-01 HS-SE-10 HS-FO-10 It was agreed that freespace optical interconnect should be considered as being outside the scope of the standard. It was agreed that AC's proposed text and diagrams on the relationships and obligations between the layers of the protocol should be developed further. AC questioned whether an option for the packet format should provide for CRC (or some similar form of error checking). PT pointed out the difficulties of implementation with wormhole routing, and also considered that this form of checking is better performed at the transaction level on an end-to-end basis - which is perfectly feasible with the standard as currently proposed. It was agreed that this is an important topic, and that an informative annex should give recommendations on error handling. 5, 5.1, 5.1.1 DS-Links - electrical The proposed method of allowing the operating speed to be a matter for system design and configuration was acccepted with minor modifications. It was agreed that figure 3 needs to be related to bit-time, rather than being specified in ns. An open issue is whether the transmission line for DS-SE should be specified at 100 Ohms or 50 Ohms. The power saving provided by 100 Ohms is significant for router chips (which may drive, say, 32 links). However, board and backplane engineering is also a significant issue. This is considered primarily a cost issue, rather than a performance issue. Is there a big cost differential? Are there significant cost issues in mixing the two on a single board (DS-DE is specified at 100 Ohms)? At short distances, it does not matter as transmission line effects are negligable. Feedback is sought from all members and interested parties on the costs and factors involved. Working papers on this issue are sought. ACTION ALL. It was agreed that it is important to fix a value. (The remainder of 5.1 was not reviewed, to save time) 5.2, 5.2.1 DS-Link - off board electrical It was agreed to specify this as pseudo ECL (ECL levels shifted by 5 volts). Input common mode voltage was provisionally specified at min -1.2V and max 7.2V. New parameters for output high and output low (typ 4.2V and 3.2V respectively) were agreed to be added. 5.2.3 DS-Link off board copper connector RB showed the latest connectors and related developments. The latching system is much improved and this was warmly welcomed. He showed work on constructing multiple connectors. This was felt to be very useful, and the specification on connector spacing was tightened up to allow for this. The connector characteristics were reviewed in detail, and various changes made. It was agreed to relegate the pcb specifications to an informative annex. It was agreed to add specifications for both min and max withdrawal force for both an unlatched and a latched connector. Specifications should be added for crosstalk, contact- contact and contact-shielding impedance, together with a description of an appropriate test environment. PM agreed to search out relevant internal evaluation work, and, given confidentiality constraints can be relaxed, release results to the group to assist in setting appropriate values. Specifications should be added or implied for vibration and shock. RB agreed to check the relevant IEC specifications to see if this can be done by simple cross- reference. A specification should be added for the tensile strength of the cable/connector clamp (this should be simply `greater than the maximum latched withdrawal force') There was significant discussion on the relationship between the specifications for the buffers to the connector, the connectors themselves, and the cables. It was agreed that the specifications for each of these should be set so that any combination can be guaranteed to work. It was proposed that a statement should be included to the effect that a connector shall not be positioned more than 3 cm track distance (pin to pin) from the corresponding buffers unless matched to 100 Ohms. RB agreed to prepare new diagrams (yet again) for the connector, showing only the relevant detail. He would minimise the number of separate diagrams, consistent with clear description at half-page diagram size. This would mean one or two diagrams for each of the male connector, female connector, and pcb layout (for the informative annex). 5.1.2 DS-Link Cable It was agreed that the electrical properties, plus mechanical performance are important for the standard, but that the detail on construction should be relegated to an informative annex as an example of a cable satisfying the specification. The electrical characteristics were reviewed in detail, and various minor changes were made. A specification should be added for EMC on the connector and on the cable. The issue of a combined budget (connector-cable-connector) for various electrical properties (for example, attenuation) was discussed further. It was agreed that this should be specified separately for each, but in such a way that the buffer-to-buffer requirements would be automatically satisfied by any conformant connector with any conformant cable. The wiring schedule for the cable/connector should be revised to omit the conductor details (pair number and colour!). 5.3, 5.3.1 DS-Link Optical connector CWS tabled the information received from Sylvain Paineau (AT&T MAC connector). It was suggested that the same connector outer physical format should be used for optic as for copper. SP and RB would discuss further with Gene Schramm, and this topic could be taken further at the Boston meeting. 5.5 DS-Link optical character encoding It was noted that this information is very preliminary, and may not represent the latest developments in detail - PT undertook to review it in preparation for the next draft. 6.1 HS link Copper connector and cable. The micropackage section should be relegated to an informative annex It was agreed that it is important at gigabaud speeds to specify pcb properties (attenuation, vias, etc) It was noted that the attenuation budget implies a cable length of about 6 metres (with the mini-coax specified). Considerations of cable properties to be included in the standard, and those given in an informative annex, are the same as for DS-Links. RB/AC reported that Harting and Bull are to have a meeting in the near future to progress the specification of a suitable connector. It was agreed that it would be desirable to carry over the mechanical properties of the connector from the work on the DS connector. A new issue was raised as to whether there should be a differential standard for HS links, in order to raise the transmission distance without incurring the expense of fibre. AC agreed to investigate whether there are suitable buffers. 6.2. HS link optical interconnection. The issue was discussed as to whether we need multiple fibres between systems, and so should specify a multiple fibre connector. It was felt that in a system, most of the connections would be copper, and short distance. Longer distance, box to box (eg down corridors) would be often implemented by a single connection (one fibre in each direction). However, further comments on this issue are welcome. 5 Next draft It was agreed that AC should work with PT and CWS by visiting Bristol during the week of 29th September and prepare a new draft, taking into account inputs received from the various actions untertaken by WG membrs during the meeting. It was further agreed that this should then be distributed immediately to the entire roster list, and reviewed at the Boston meeting. 5 Open issues The list of open issues was not reviewed in detail, other than most of the topics had already been covered in discussion, as noted above. A list of current open issues should be drawn up and distributed C Whitby-Strevens - Chairman 3rd September 1993