Date: 9 Sep 97 13:51:49 -0700 Subject: Missing Minutes, Even Ours ... From: "Churchill, Richard" To: "Wooten,David" , 'mike@fireflyinc.com', 'colinws@bristol.st.com', 'ehannah@mipos2.sc.intel.com' X-Mailer: Cyberdog/2.0 MIMEEnabled: Y MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="Cyberdog-MixedBoundary-000E88B8" Content-Transfer-Encoding: 7bit --Cyberdog-MixedBoundary-000E88B8 Content-Type: text/enriched; charset=macintosh Content-Transfer-Encoding: 8bit X-Fontfamily: Monaco X-Fontsize: 9 Gentlemen, I noticed on examining the FTP site index.html and actual directory for p1394b that minutes for the May 5 long haul and general meeting, August 7, 1997 high-speed long-haul meeting, and August 8, 1997 general meeting are not listed. I know I sent these out a while back (in the case of the May meeting I did so in May), but, to ensure that these are available, here are those minutes again. Sincerely, Richard Churchill, Adv. Portable PC Arch., Compaq Computer Corporation, P.O. Box 692000, ms120307 Houston, TX, 77269-2000 20555 St. Hwy. 249, ms120307, Houston, TX, 77070 richard.churchill@compaq.com phone: (281)514-6984 fax: (281)518-5324 GenevaP1394b Working Group Meeting, May 5&6, 1997, Santa Clara, CA, Embassy Suites Hotel Michael Teener, Chair, absent at start Eric Hannah, Editor, absent at start Richard Churchill, secretary Colin Whitby-Strevens, Long-haul task group chair <<<<<<<<<<<< PART A (Long-haul) OF MEETING >>>>>> (The spelling of "fibre" is in deference to Colin.) I. Administrivia Meeting called to order by Colin Whitby-Strevens Note made of traffic over Santa Cruz Mountains, delaying Mike Colin reported addition of long-haul duties at Eindhoven, as agreed by Mike Teener. Questions regarding the addition of task were raised, and were answered by Colin, stating that 1394b now has this additional work with the expectation that it might later be spun off into a seperate effort. Agenda 1. Introductions 2. Approval of Agenda 3. Background 4. Review of objectives and guidelines 5. Proposed model 6. Review requirements 7. Contributions David Smith (Honeywell) - VCSEL technology for optical fibre Patrick Yu (NEC) - POF 8. Discussion - timescales, etc. 9. Action items 10. AOB 1. Introductions around .... 2. Agenda was agreed. 3. Background Mike Teener explained reason for a two-day meeting, rather than the traditional one-day: Expansion of task to long-haul/gigabit operation. Task group will assume responsibility for bit-push, remainder of group with worry about SW, silicon, etc. ... Future problem getting and keeping meetings coordinated. Issue regarding appropriateness of addressing the 100 Mb/s UTP5 problem -- would require changes to PAR, etc. "PMD for Long Distance 1394 Kick-off meeting" by Colin follows VESA HN Committee selected 1394 as cluster interconnect VESA HN selected 1394 as basis of home backbone needs an appropirate physical layer needs protocols for IP, etc. PHY working group within VESA HN to develop technology demo as first step towards spec part of ... ..... 4. Review of objectives and guidelines -- continuation of presentation by Colin Objectives Evaluate and refine the requirements Consider the use of Optical Fibre and UTP transmission media Consider the long distance transmission at all 1394 cable environment speeds, ... Prioritise the determination of a medium recommendation for new domestic installations [*** Considerable discussion seeking to clarify issues, such as the requirement for UTP. Mike made the point that this arguement for UTP is spurious, since the requirements will be different in 3 years, and fiber is easier to install than coax ... Dick Scheel pointed out the "fear" factor associated with fiber, and that this is an education problem as much as ahything ... ***] Guidelines 5. Proposed Model(s) -- continuation of presentation by Colin Model 1 [drawing ...] A new reference point is defined, called PMII (Physical Media Independent Interface). The current P1394b work defines everything towards the PHY from this interface, the new PMD subgroup defines everything towards the medium from this interface. Model 2 - There is one specification for PMII, but a range of specifications for 1394b-n (n is used to distinguish one from another) - The PMD-n transceivers are unintelligent and non-configurable. The role of a transceiver is to convert between the signals provided by the PMII and a suitable format for long-distance transmission. - The PMII electrical specificaiton is the specification currently being defined within 1394b for S800 and upwards short copper links. In this casel, the PMD transceiver is null, and the PMII = 1394b-N, (where N is the value of n which identifies this interface. - The PMII may or may not be connectorised. When connectorised, it uses the standard 1394 connector. [*** Discussion of assorted implementation issues, including cost. Point raised that competing technologies have had no initial discussion of or concern for cost -- which is opposite of "traditional" 1394 approach. ***] Model 3 The following packagin options are permitted (non-exhaustive list):- - The external interface of the device is the current 1394 connector. A short 1394 copper cable leads to a dongle, which converts to an appropriate optical interface, and a short optical cable is termin- ated in an optical connector. - As above, with the optical cable replaced by a long-haul copper cable (e.g. UTP-5). - The external interface of the device is the current 1394 connector. a 1394 copper cable leads to an active wall plate with another 1394 connector, the transceiver, and connections to installed fibre (or long-haul copper). - The external interface of the device is an appropriate optical or long-haul copper connector. The long reach transceiver is internal to the device, the PMII is not connectorised, and may be subsumed within an integrated implementation. The speed capabilities of the PHY may exceed the speed capabilities of the PMD transceiver. The PHY needs to discover the bandwidth capabil- ities of the channel with no user involvement. The start-up procedure is therefore defined across the PMII. [*** Discussion of assorted issues ... Mention made of problems involved in creating any standard with active wall-plates, which approaches the impossible. Assorted discussion of implementations, costs, etc. .... ***] 6. Review requirements -- Continuation of presentation by Colin PMD Proposed Requirements - 50m reach (100m preferred) per hop - determine worst-case delay - UTP-5 and Optical Fibres - minimise the differences between these two media - low complexity implementation - Facilitate FCC Class B emissions compliance - P1394b above the PMD layers - same tree-ID algorithms, self-ID algorithms - fully interoperable with current 1394 - no bus bridging required - Amateur installable - Installation guidelines, installation test, ... [*** VESA had a considerable discussion of requirements -- the round-about paths of high-end installations cna easily reach substantial one-hop reaches. Surveys indicated 50m was rather short for the US market, while 100m covers the vast majority of the US housing market. Every meter over 50m adds a percentage of the US home market. 50m was only acceptable if necessary, but leaves out a significant portion of the market (comment by Dick Scheel) Discussion of point by others regard- ing requirements and trade-offs. POF bandwidth and reach issues were discussed. 100m requirement suggested, but 70 to 100m offered as a good compromise. Discussion of the UTP-5 and Optical Fibre delta minimization discussed, with point made that the minimal delta was with respect to the i/f. Opinion expressed regarding low-priority for UTP support, with support in fact permitted as incidental, but not as an objective. Colin: "Cost is King!" -- "... Must decide what priorities are ..." Added P1394a at start of bullet 4 before "P1394b ..." Question raised regarding bridging ... Is it a priority that no bridging be required? (Yes!) Lasers must be class 1 EYE-SAFE (IEC 825-1 and CDRM (r-DA)) for regu- latory purposes in order to be amateur installable. David Wooten raised issue of the incompatibility of the requirements, stating that the simple, low-cost objective is not compatible with the requirement to support "all known media at all known data rates." -- LEDs can get us to about 300 Mbaud (some to 600) -- POF gets us to about 150 Mbaud -- laser from 300/600 Mbaud to 1+Gbaud Question/Comment that we need to be aware that we should consider the availability of technologies --- 300 Mbaud may be possible with LEDs, but is anyone actually selling such a solution? Might we practically be limited to 155 Mbaud by the vendors when using LEDs? "It has not been demonstrated that POF is necessarily cheaper than glass fibres ...." Additional discussion about how to approach the problem ... "encap- sulated" within the cable, or within the wall seems the simplest ... We should look at the state of research, and see how we can interplay to our mutual benefit to reach a better cost-point ... Strike the distinction between MMF and SMF, which gets us to seven options for prioritization ... Questions regarding model ... Model must support incremental upgrade. ***] Votes for priority ... (Each person present was permitted a single vote in each of the two priority levels.) I II Speed & Technology -- -- --------------------------------------------------- 0 2 S100/UTP 0 5 S100/Fibre 1 1 S200/Fibre 1 2 S400/Fibre 11 0 S800/Fibre 7 0 S1600/Fibre 0 8 S3200/Fibre (This vote implies at least two more abstentions in the "second priority" vote than in the "first priority" vote.) (1) lowest cost, and (2) S800 and upwards (???) Should we support the same layers "above" the PMD? (Shall the encoding schemes, etc., used be identical for both low-speed and high-speed solutions?) Motion moved by Richard Churchill, and seconded by Colin Whitby-Strevens. The vote was 20 yeas versus 2 nays, and the motion carried. 7. Contributions David Smith (Honeywell) - Vcsel (vertical cavity surface emitting laser) technology for optical fibre Run on much lower current ... Testable on wafer ... significantly lower cost than edge emitting lasers ... Central active region between p- and n-mirror stacks ... Made with multiple quantum wells ... AlGaAs, AlAs, GaAs, etc., used Comparison of properties ... Sub milliamp thresholds are now being displayed ... Encapsulated, Class IIIb can be used in a Class I manner ... Modulation current 3-10 mA, CD lasers >10 mA Spectral Width 0.5nm, vs CD laser 9 nm, and LED 50-200 nm ... Patrick Yu (NEC) - POF - Reiteration of parts of his Eindhoven presentation ... Goal of 200 Mb/s this Fall ... 1394 POF requires "fixes" for (1) Arbitration, (2) Self_ID, and (3) encoding (to use 4B5B). One 70m hop over POF roughly equivalent to 3 hops in "traditional" 1394 -- roughly 350 ns -- exclusive of the implied outer- edge repeaters [*** Is the attenuation of the POF (10's of dB/100m) acceptable, given con- sequences regarding laser power vs. cost and eye-safety? Will the POF, GI, Gb capable, be acceptable to the high-speed 1394 folks? Indications are that this is not presently available, and there is no reliable knowledge regarding when it will be. Vendors must be checked ... Possible Action Item? Who? ***] 8. Discussion - timescales, etc. [*** Does the model presented imply that we will see VCRs, PCs, etc., with optical connectors? Quoth Mike, "Let's not worry about that now. If we work on the technologies, and make sure we can all do them, maybe the connector problem will go away. ... Discussion of cost of media vs. cost of installation vs. cost of "finishing" (termination, photonics, etc.). ***] 9. Action items < Colin is "nominated" to be chair of high-speed long-haul, and Taka Fujimori to be chair of the "lowest cost" effort. Colin passed sign-up sheets for both groups 10. AOB <<<<<<<<<<<< END OF PART A (Long-haul) OF MEETING >>>>>> <<<<<<<<<<<< START OF GENERAL P139B SESSION >>>>>> Presentation by Bill Northey (Berg) updating work done on improvements the existing six-pin connector and socket. Data rate improvements by geometry changes indicates 3 Gb/s data rate is achievable. see previous meetings presentation ... Offered to share simulation package data with any who are interested. Initial samples in July, with production in October. Summer Schedule Discussion Early June meeting in Seattle proposed, with a skip for July ... Suggestion by East Coast folks that meeting NOT be on Monday, so that they will not be forced to travel on the weekend. General agreement on week of June 9, possibly in Seattle. Proposal for a meeting in middle of September, with Colin W-Strevens suggesting Bristol, with a two-day meeting, plus suggestion that Peter be contacted regarding holding a P1394a meeting there at an adjoining time. October meeting in conjunction with TA in Phoenix. IBM presentation on 8B/10B. IBM's 8B/10B coding scheme is available for licensing for 1394b, with contact information, etc., as follows ... Al Torressen, (914)742-6275, TORRESS@US.IBM.COM, for a $5,000.00 license of the 8B/10B code. (This constitutes a completed Action Item from the previous meeting!) <<<<<<<<<<<< Recess for night at 4:40 PM, resuming Tuesday, at 9:00 AM. >>>>>> <<<<<<<<<<<< Meeting resumed at 9:35 AM, Tuesday May 6. >>>>>> Mike Teener was again delayed by traffic. AGENDA 1. Approval of Agenda 2. Introductions 3. IBM License 4. Presentations a. Start-up b. Codings Keith Heilman Taka Fujimori Jerry Hauck Alistair c. Electricals Karl Nakamura 1. Approval of (provisional) Agenda Done. 2. Introductions 3. IBM License Done. Same as previous day. 4. Presentations a. Start-up, Colin Whitby-Strevens (SGS-Thomson Micro.) (Six page document, titled "P1394b Start-up," distributed, with authors listed as "Colin Whitby-Strevens and friends.") Introduction - Start-up procedure for Beta-capable links - Independent of the coding shceme used - some requirements on such coding schemes - Goal is to start up in Beta mode if both ends of a connected link are Beta capable, otherwise start-up in DS-Link mode. - Normal bus reset, Tree_ID, Self_ID, and operation will follow AFTER this start-up procedure. - The techniques aims to be as sympathetic with the current 1394 philosophy as possible. Requirements - To start-up as an ordinary DS link if either end is not Beta capable - To start up in Beta mode if both ends are Beta capable - To start up in Geta mode at the maximum baud rate which is acceptable to both ends (this assumes a single operating speed in terms of baud, but with some form of padding to support lower speed packets.) - To support the higher speed copper links being defined with P1394b including speed determinations .... - - - - Sleep mode [Mike Teener arrives] [Physical Medium Dependent -- PMD] Model 1 diagram from previous day's presentation re-presented. Models 1-3 from previous day's presentation re-presented. Procedure - use of TPA and TPB - 1394-1995 uses both the TPA and TPB pair for transmission and reception of data, as well as for digital DC arbitration signalling - Common mode signalling is performed with respect to a bias volt. generated on TPA. This voltage is sensed on TPB as connection indication - current is also drawn at TPB which is sensed as a voltage drop at TPA for signalling. - It is required tht the PMII supports these facilities unchanged when the PHY is operating as a DS Link - In Beta mode, data is transmitted on TPB and received on TPA - the potential use for optical transmission implies either (i) there is no DC signalling across the PMII (ii) ... Start-up procedure, main states - The power-on start-up procedure for a Beta-capable PHY proceeds through the following states 1. Connection status: 2. Speed determination 3. Tree_ID, et. - After state 2, normal operation of the link is continuous - A procedure is defined for recovery from loss of synchronization - A procedure is defined to maintain a conection in a low-power mode where transmission is not continuous - Connection status - 1 - This state determines whether - - - - The use of TpBias is restricted to its use as specified for 1394-1995 DS Link signalling. A beta ... Connection status - 2 - On power on, a Beta capable PHY should wait for the TpBias debounce time, and then check for TpBias (on TPB) on each of its ports. - If TpBias is sensed, this implies that the port at the "other" end is a DS Link only device, and is enabled. The PHY generates TpBias on TPA, and the port starts up in DS mode as per P1394a (reset, Tree_ID, Self_ID, ...) - If TpBias is not sensed, then a sensing tone is transmitted (e.g. 125 MHz (62.5 MHz?)), and listened for. - If, at any time, TpBias is sensed, then this procedure is abandoned (this will occur if a DS-only device is switched on and its port enabled just after the Beta-capable device is turned on). - The tone is transmitted for a defined period of time - long enough to overcome the time constands in the PMD. 5.0 usec is proposed, but needs to be balidated by the PMD group as being sufficient to overcome the time constants in all anticipated PMDs. Connection status - 3 - If a tone is received, then this confirms that the "other" end is Beta capable. The PHYs at both ends enter speed determination state. - Care must be taken when specifying the frequency of the tone that will be low enough in frequency to pass through the slowest PMD, but .... Speed determination - 1 - This is largely still open, and requires further study. - The pHYs can exchange information giving their maximum operating speed, but this does not deal with the problem of a slower transceiver. - A possible method which inclues a crude transceiver test is described below - Three symbol sequences are defined -S1, S2, and S3 - For ease of understanding, consider each sequence as 2/4 10- bit symbols. - The choice of symbols depends on the coding scheme to be adopted. Each sequence has the property that the PHY can per- form .... - .... Speed determination - 2 - The sequence S! is transmitted repeatedly at the S100 equivalent rate (122.88 MBaud) - Simultaneously, the sequence is listened for. - The receiver performans syumbol alignment until it can receive the sequence successfully - The exchange of S1 continues for long enough .... - ... Speed determination - 3 - If the PHY is capable of a higher speed, it then transmits S3 repeatedly, otherwise it continues to transmit S2. - If it is capable of a higher speed, and it receives the S3 sequence, then it starts transmit the S1 sequence at the S200 equivalent rate, and attempts to lock reception at this rate. - Upon successful receptoin for a defined period of time, the PHY confirms receipt by changing the transmitted pattern to S2. The process repeats to the next higher speed. - In the event that transmission is not successfully received, then the PHY reverts to the lower speed. - Care needs to be taken to deal with the case when the trans- mission is OK in one direction, but not the other. - The result of this procedure is bi-directional transmission at the maximum rate acceptable to the PHYs, the transceivers and the medium. Bit alignment and symbol alignment is achieved - Upon completion of the procedure, the PHY commences to transmit. - The PHY also sets the Bias bit (simulating TpBias for the PHY-LINK interface), with the consequent implications for the Con bit. [Colin extended an apology for confusing the terms "link" and "port" in this presentation. Reader should take care to observe the usage in interpreting the presentation.] [*** Extended discussion of speed determination and reporting occurred here, with various questions raised, and in turn answered by assorted other attendees. ***] Choice of Sequences - The properties required of a sequence are - Each sequence should be DC balanced - Each sequence must provide suitable patterns for symbol alignment - IBM 8B/10B - Fibre Channel OLS (for S1), LR (for S2) and LRR (for S3) - similar to the fibre channel start-up procedure (but Fibre Channel des not have to provide a speed determination function) - Coles 8B/10B code (SC1 .... - ... - ... - ... - ... Await Connection - The beta capable PHY has received neither TpBias, nor a tone - It transmits a tone for a period of n usecs (n = 5?) - This is repeated at intervals of 60n usecs until either a tone is detected or TpBias is detected on TPB - If a tone is heard, then the speed determination procedure is entered - If TpBias is detected, then TpBias is generated on TPA, and start-up is as per 1394-1995 - This scheme avoids excessive energy being transmitted into an unterminated cable. - This would be necessary for a UTP 1394 link, but conuld also be a sensible move for any 1394b copper link. - Also consider use of "sleep" modes, and the need to be able to detect a wake-up signal without requiring the PHY to be powered all the time. - This may be achievable either by - a "signal detect" signal from the transceiver (communicated to the PHY by using some form of TpBias-type signal. - By having a low level active circuit which can respond to the reception of the tone and wake up the rest of the PHY - The second option is preferable. ???? - This is a topic for further study - However, a good starting point is the Fibre Channel state machine - provides hysteresis on the detection of an error (e.g. an invalid symbol, or excessive running disparity) - Synch is maintained unless a rapid sequence of errors is detected. - The PHY transmits an indication of loss of synch. - IBM 8B/10B code: this could be the NOS primitive sequence - Coles 8B/10B code: this could be the S1 sequence - The PHY enters the speed determination state on loss of synch. and on receipt of the loss of synch indication - the need to consider the seret behaviour on loss of synch. Summary - Start-up takes place BEFORE reset, Tree_ID, Self_ID, etc., - arbitration state machine is unchanged from 1394-1995 - Beta capable PHYs adapt _ _ [*** Extend discussion on numerous topics ... Suggestion of creating an "Open Issues" list as a "living document." ***] b. Coding -- Serial Bus Packet Format, Keith Heilman presentation (material has been on reflector) foil 1, diagrams, 1394-1995 start-stop to 1394b continuous 1394B Transmission Speed table ... 1394B Data Format table & brief comments ... 1394B Arbitration Format 2 Byte ordered sets for Arbitration signals ... (Encoding remapped) [ R | SPD |B1 |B0 |A1 |A0 ] [ | | | | | | | ] foil 5, table, 1394-1995 Arbitration encoding for IBM 8B/10B [*** Mike Teener accepts homework of developing method of using the now full-duplex connections to remove all gaps by performing arbitration in parallel with data transmission .... ***] <<<<<<<<<<<< RECESS FOR LUNCH AT 12:06 PM >>>>>> <<<<<<<<<<<< RESUME AT 1:20 PM >>>>>> -- "POF/UTP 1394 Long (100m), Based on 4B/5B Code" by Taka Fujimori Room-to-room connection via POF/UTP 1394 diagram Characteristics of POF/UTP 1394 PHY 1. One hop length of POF 70m @ 100/200 Mbps UTP Cat 5 100m @ 100 Mbps 2. Coding is the same as 100Base-T - 4B/5B Code Tech. is already proven - O/E, E/O Link accepts 10% unbalance with absolutely no additional cost 3. Passes FCC Class B 4. Cost is minimal 5. Low cost 650nm LD for POF 6. ATM Forum has passed SI (Step Index) POF 155 Mbps 50m with 2 splicing points UTP/POF 1394 Repeater Block Diagram diagram Valid 4B/5B Code table 4B/5B Code Mapping to 1394 table .... How Packets Look Like S200 Long PHY. .... ... S400 Long PHY. .... ... Conclusion 1. Mapping 1394 control onto 4B/5B is Complete Max Speed Signalling is unlimited: S800, S1600 ... 2. Sony & Hyundai submitted a Joint Proposal to DAVIC Hyundai submits IP (4B/5B) Statement 3. Proven by 100Base-T Ethernet, Minimal Cost Pass FCC Class B 4. Verilog Code will be available. -- HP 8B/10B Block Code, by Alistair Coles (Similar to previous HP presentation on HP/Coles 8B/10B) Spectrum of HP 8B/10B Block Code graph of spectral character of unencoded, MLT3 and HP 8B/10B codes. 8B/10B Block Code: Operation - Data: - Data codetable has 256 codewords: 111 have zero disparity (i.e. equal number of 1s and 0s - Encoder updates the running digital sum (rds) at the end of transmitted ... - - - - Control (Arbitration) - Control codetable has 8 codewords: all have zero disparity, and are never inverted. - These 8 codewords are Hamming distance 2 from ... - ... Block Diagram: Data/Control Paths diagram - 8B/10B coder uses separate code tables for data and control: 256 data codeowrds, 8 control codewords. - ... - ... ..... Synchronization - two codewords (SC1 = 1010101010 and SC2 = 0101010101) areprovided for synchronization - Local End of a link begins training by repeating the pattern {SC1, SC1, SC2, SC2}. Far end will always se an occurence of SC1 or SC2, even when not synchronized - Far end responds with the pattern {SC1, SC1, SC2, SC2}. - Far end uses transition between SC1 and SC2 to achieve codeword alignment. - Local end uses transition between SC1 and SC2 to achieve codeword alignment. - When codeword alignment has been achieved, repeating pattern is ... - ... [*** No present plans by HP to pursue a patent on this coding scheme. ***] 1394B: Control Signal Mapping - If 1394B PHY is to support PMD: At all cases consideration should be given to control signal timing at lower rates. e.g. At S100, two codewords (K28.5 ... [*** Discussion of Scrambling, Resynchronization, etc., ... Question regarding utility of Hamming distances of two between data and control characters, and between control characters. Proposal to be drafted by Mike Teener for arbitration w/o gap. 2 error conditions: GNT to device w/ no request and no GNT (fairness will take care of). MT will come up w/ simple codes and propose for next mtg. Gate count for 8B/10B IBM ~ 1K gates Coles (HP) 8B/10B coding for 100 Mbit PHY repeater in alterra 115K gates for total. For the coding, no numbers have been generted. AR: Get the verilog w/ and w/out the scambler. Look at publishing the verilog Alistair Cole The URL for the HP codes www.fireflyinc.com ***] -- Bit-ordering for IBM 8B/10B Coding, by Jerry Hauck This material was supposed to be distributed via the reflector, but had not been seen as of presentation. Presentation was informal, with intention of leading to reflector discussion. Burst errors caught by CRC up to 32-bits, but 8B/10B produces a 40-bit symbol set from 32-bit object, which may be a burst error, making for a 40-bit burst error that cannot be guaranteed to be caught by present CRC. Clarifying memo required to be consistent with present conclusion, Manner of bit-stuffing will need to be considered ... c. Electricals Changes in Gigabit Ethernet, by Karl Nakamura Brief comments on changes to Gigabit Ethernet, and finalization thereof. [*** Invitation to all from Mike Teener -- What signal levels should we be using for IBM 8B/10B coding, given our known characteristics (present cable and present or presume improved connectors ...), and what should the input sensitivities be? What are the eye- opening requirements? Karl has some information on this ... ***] Action Items .... Teener -- background arbitration ... Taka Fujimori -- Finding a common speed between connected PHYs Jerry Hauck -- clarification of legendary CRC, transmission ordering, etc. document on the reflector Eric Hannah and Karl Nakamura -- signalling requirements, etc., for 1394b All -- Comments and proposals based upon Colin's Start-up presentation/ proposal ... ============================================================================= Attendees: 1 Colin Whitby-Strevens SGS-Thomson +441454611500 colinws@bristol.st.com 2 Richard L. Churchill Compaq 281-514-6984 richardc@bangate.compaq.com 3 David R. Wooten Compaq 281-518-7231 davidw@bangate.compaq.com 4 Jerry Hauck Intel 408-765-5528 jerry_hauck@ccm.sc.intel.com 5 Steve Midford Intel 408-765-8370 steve_midford@ccm.sc.intel.com 6 Karl Nakamura LSI Logic 408-433-4516 karln@lsil.com 7 Taka Fujimori Sony +81-3-5488-6353 fujimori@arch.sony.co.jp 8 Bill Prouty HP 916-786-4631 billp@hprnd.rose.hp.com 9 Sumihiro Okawa Sony +81-466-30-4050 okawa@sm.sony.co.jp 10 Hiroshi Takizuka Sony +81-466-30-4050 takizuka@sm.sony.co.jp 11 David Smith Honeywell 972-470-4402 dsmith@micro.honeywell.com 12 Dick Scheel Sony 408-955-4295 dicks@lsi.sel.sony.com 13 Joel Urban Harting 847-741-1500 [email address not given] 14 D. Desourteaux Harting [no phone number given] d.desourteaux.harting@w????.fr [correction requested] 15 Kazuyuki Abe NEC System Lab 415-528-5904 kabe@sj-pceg.ccgw.nec.com 16 Peter Teng S3, Inc. 408-588-8111 pteng@mail.com 17 Ming Qu National Semiconductor 408-721-8922 mqu@rockie.nsc.com 18 Raghunath Cherukuri Texas Instruments 972-480-3384 crn@msg.ti.com 19 Claude Huss Matsushita E.W. 408-433-3386 claude@mew.com 20 Sunny Lam IBM 512-838-6276 sunnyl@us.ibm.com 21 Shinichiro Matsui NEC [no phone number given] s-matsui@nims.nec.co.jp 22 Robbie Shergill National Semiconductor 408-721-7959 rss@berlioz.nsc.com 23 Dao-long Chen Symbios Logic 970-223-5100 x9461 dao-long.chen@symbios.com 24 Bradley N. Saunders Rockwell 714-221-6513 bradley.saunders@nb.rockwell.com 25 Magami Tsugita NEC [no phone number given] tsugita@lsi.ting.nec.co.jp 26 Hiroaki Kobayashi NEC [no phone number given] kobaya@nims.nec.co.jp 27 Hideki Negishi NEC [no phone number given] negishi@kansai.nims.nec.co.jp 28 Keith Heilmann IBM 914-892-2413 heilmann@vnet.ibm.com 29 Michael Johas Teener Firefly 408-461-4901 mike@fireflyinc.com 30 Cyrus Momeni Cirrus Logic 510-624-7061 cyrus@corp.cirrus.com 31 Charles Brill Amp 717-592-6198 cebrill@amp.com 32 Bijit Patel Lucent 610-712-2188 patelb@lucent.com 33 Eric Bergles CEL 408-588-2216 ebergles@cel.com 34 Steve Bard Intel 503-264-2923 steve.bard@ccm.jf.intel.com 35 Bill Northey Berg 717-938-2119 northewa@bergelect.com 36 Patrick Yu NEC Electronics 408-588-5436 patrick_yu@el.nec.com 37 Jonathan Buck Amp Inc. 408-725-4923 jon.buck@amp.com 38 Mike Gardner Molex, Inc. 408-946-4700 mgardner@molex.com 39 Ketan Patel Molex, Inc. 630-527-4047 kpatel@molex.com 40 Max Bassler Molex 630-527-4490 mbassler@molex.com 41 Rajiv Choudhary Intel 503-264-8480 rajiv_choudhary@ccm.jf.intel.com 42 Jack Merrow Leviton Telcom 206-486-2222 jmerrow@leviton-telcom.com 43 Michael Nguyen Fujitsu 408-894-3542 michael.nguyen@fcpa.fujitsu.com 44 Firooz Farhoomand Matsushita/Panasonic 408-945-5625 firoozf@netcom.com 45 Alistair Coles HP Labs +44-117-922-8750 anc@hplb.hpl.hp.com 46 Mike Brown Intel 602-554-3713 mike_brown@ccm.fm.intel.com 47 Heather Lane SRTI 415-988-4812 heather@srti.com 48 Gene Matter Intel 916-356-4597 gene_p_matter@ccm.fm.intel.com 49 Nobuo Furuya NEC Electronics 408-588-5414 nobuo_furuya@el.nec.com 50 Benjamin Pan S3 Inc. 408-588-8379 bpan@s3.com 51 Kugao Ohuchi NEC Electronics 408-588-5503 Kugao_Ohuchi@el.nec.com 52 Khorvash Sefidvash Symbios Logic 714-474-7095 khorvash.sefidvash@symbios.com 53 Mike Winchell Symbios Logic 970-225-4807 mike_winchell@symbios.com ----------------------------------------------------------------------------- S100 Long Haul Signees: 1 Colin Whitby-Strevens SGS-Thomson +441454611500 colinws@bristol.st.com 2 Charles Brill AMP 717-592-6198 cebrill@amp.com 3 Bijit Patel Lucent 610-712-2188 patelb@lucent.com 4 Eric Bergles CEL 408-588-2216 ebergles@cel.com 5 Mike Brown Intel 602-554-3713 mike_brown@ccm.ch.intel.com 6 Max Bassler Molex 630-527-4490 mbassler@molex.com 7 Gene Matter Intel 916-356-4547 gene_p_matter@ccm.fm.intel.com 8 David James Express Associates 415-494-0926 dvj@ibm.net 9 Patrick Yu NEC 408-588-5436 pyu@el.nec.com 10 Raghunath Cherukuri Texas Instruments 972-480-3384 crn@msg.ti.com 11 Rajiv Choudhary Intel 503-264-8480 rajiv_choudhary@ccm.jf.intel.com 12 Sunny Lam IBM 512-838-6276 sunnyl@us.ibm.com 13 Shinichiro Matsui NEC [no phone number given] s-matsui@nims.nec.co.jp 14 Dao-long Chen Symbios Logic 970-223-5100 x9461 dao-long.chen@symbios.com 15 Hiroaki Kobayashi NEC [no phone number given] kobaya@nims.nec.co.jp 16 Hideki Negishi NEC [no phone number given] negishi@kansai.nims.nec.co.jp 17 Magami Tsugita NEC [no phone number given] tsugita@lsi.ting.nec.co.jp 18 Mike Winchell Symbios Logic 970-225-4807 mike_winchell@symbios.com 19 Dick Scheel Sony 408-955-4295 dicks@lsi.sel.sony.com 20 Taka Fujimori Sony +81-3-5488-6353 fujimori@arch.sony.co.jp 21 Sumihiro Okawa Sony +81-466-30-4050 okawa@sm.sony.co.jp 22 Hiroshi Takizuka Sony +81-466-30-4050 takizuka@sm.sony.co.jp 23 Alistair Coles HP Labs +44-117-922-8750 anc@hplb.hpl.hp.com 24 Eric Hannah Intel 408-765-4441 eric_hannah@ccm.sc.intel.com +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ S800 Long Haul Signees: 1 Colin Whitby-Strevens SGS-Thomson +441454611500 colinws@bristol.st.com 2 Charles Brill Amp 717-592-6198 cebrill@amp.com 3 Bijit Patel Lucent 610-712-2188 patelb@lucent.com 4 Mike Brown Intel 602-554-3713 mike_brown@ccm.ch.intel.com 5 Gene Matter Intel 916-356-4547 gene_p_matter@ccm.fm.intel.com 6 David James Express Associates 415-494-0926 dvj@ibm.net 7 Patrick Yu NEC 408-588-5436 pyu@el.nec.com 8 Raghunath Cherukuri Texas Instruments 972-480-3384 crn@msg.ti.com 9 Rajiv Choudhary Intel 503-264-8480 rajiv_choudhary@ccm.jf.intel.com 10 Sunny Lam IBM 512-838-6276 sunnyl@us.ibm.com 11 Shinichiro Matsui NEC [no phone number given] s-matsui@nims.nec.co.jp 12 Dao-long Chen Symbios Logic 970-223-5100 x9461 dao-long.chen@symbios.com 13 Bradley N. Saunders Rockwell 714-221-6513 bradley.saunders@nb.rockwell.com 14 Hiroaki Kobayashi NEC [no phone number given] kobaya@nims.nec.co.jp 15 Hideki Negishi NEC [no phone number given] negishi@kansai.nims.nec.co.jp 16 Eric Hannah Intel 408-765-4441 eric_hannah@ccm.sc.intel.com 17 Jerry Hauck Intel 408-765-5528 jerry_hauck@ccm.sc.intel.com 18 Richard L. Churchill Compaq 281-514-6984 richardc@bangate.compaq.com 19 Mike Winchell Symbios Logic 970-225-4807 mike_winchell@symbios.com 20 David R. Wooten Compaq 281-518-7231 davidw@bangate.compaq.com 21 Steve Midford Intel 408-765-8370 steve_midford@ccm.sc.intel.com 22 Taka Fujimori Sony +81-3-5488-6353 fujimori@arch.sony.co.jp 23 Sumihiro Okawa Sony +81-466-30-4050 okawa@sm.sony.co.jp 24 Hiroshi Takizuka Sony +81-466-30-4050 takizuka@sm.sony.co.jp 25 Alistair Coles HP Labs +44-117-922-8750 anc@hplb.hpl.hp.com 26 Kazuyuki Abe NEC System Lab 415-528-5904 kabe@sj-pceg.ccgw.nec.com P1394b High-speed Long-haul 1:30-~5:00 PM 7 August, 1997 Ala Moana Hotel, Honolulu, Hawaii chairman: Colin Whitby-Strevens secretary: Richard Churchill 1. Welcome and Introductions 2. Approval of Agenda 3. Approval of minutes of last meeting 4. Liaison reports 4.1 Report from S100 PMD Task Group 4.2 Report from long-haul PMD, 8/6/97 5. S800 P1394b Electrical signal integrity issues Dave Brunker Eric Hannah 6. Media and Transceiver technologies 6.1 Contributions on specifications for suitable technologies 6.2 Identification of issues for further study 7. Connector Selection Tad Szostak 8. Sub-PHY root contention Dave Lafollette 9. Future plans and schedule 10. AOB 11. Adjourn 1. Introductions were made for all present, and the tentative agenda was presented. 2. The agenda was approved, with some modification (given as amended above). 3. Minutes of the previous meeting were approved without objection. 4. Liaison reports 4.1 Taka fujimori-san reported on the discussions held in the morning by the low-speed long-haul task group. That report is appended by reference to the minutes of that group. Next meeting is planned for Tokyo in September. 4.2 Colin Whitby-Strevens reported on the discussions held the previous day in the joint long-haul PMD meeting. That report is appended by reference to the minutes of that meeting. Specific items presented included the discussion of the properties of a sub-PHY, a decision to recommend to the P1394b committee of the whole to select a single optical connector, objective of a single media for all speeds (exclusive of UTP cat-5), though the option of different media for different speeds was retained as an option, and discussion of whys and wherefors of active and passive components in the wall and wall-plate. Further, considerable discussion was held upon the relative merits of the two coding schemes under con- sideration by the high- and low-speed PMD task groups. 5. S800 P1394b Electrical signal integrity issues Dave Brunker (Presentation synopsis here, but complete presentation included in minutes by reference to the soft copy available from the group ftp site.) - Molex 1394-1995 6 ckt. I/O Connector ONLY Differential Electrical Performance - Connector Only; Introduction - Data presented is for the connector only; no 1394 cable is attached. - .... - Connector Only Test Fixture - - - - - Connector only Test Fixture *** diagram *** - Connector only Test Fixture *** diagram *** - Connector only Test Fixture *** diagram *** - Connector only Test Fixture *** diagram *** Test Conditions; Differential Impedence *** table *** Load Impedance 110 Ohm differential (resistive load) Source Impedance 110 ohms differential (2x55 ohm PCB) Test Risetime (10%-90%) 26ps*, 150ps, 250ps, 500ps*, 1000ps (Equipment used was the Tektronix 11801b Digital sampling Oscilloscope with ... - SMT/Flat Differential Impedance Results *** table *** - Through Hole/Upright Differential Impedance Results *** table *** - Test Conditions; Differential Crosstalk *** table *** - SMT/FLAT Differential Crosstalk Results *** table *** - Through Hole/Upright Differential Crosstalk Results *** table *** - Test Conditions; Differential Risetime Degradation and Propagation Delay *** table *** - Test Conditions; Differential Risetime Degradation adn Propagation Delay - The risetime degradation data was calculated using the 10% to 90% levels of the input and output pulse to the connector, where: tr1connector = [(tr,DUT+test_system)e2 - (tr,test_system)e2]e1/2 (Presentation contained 40 slides, total.) < < < < Eric Hannah (material previously presented, and included by reference to the reflector.) - Spice Modeling Differential Pair Transmission Lines *** diagram *** - Differential Mode ... *** diagram *** - .... *** diagram *** - .... *** diagram *** - .... *** diagram *** - S400 ... *** diagram *** - S400 K28.5 over 20 inches + 2 connectors *** diagram *** *** diagram *** - Minimum Rise Time Study *** diagram *** *** diagram *** < < 6. Media and Transceiver technologies 6.1 Contributions on specifications for suitable technologies Dan Rausch (The was presented at the joint long-haul session the previous day. The presentation is included in these minutes by reference to the minutes of that joint session. Though not all slides were covered in that session, all were incorporated in the minutes.) < < < < Moved (Richard Churchill) Seconded (Mike Teener) that the chair of this group and the chair of the P1394b committee develop an appropriate letter to establish needed communication with TIA 41.8.2 requesting consideration of 50u fiber. Friendly amendment to include communication with TIA 41.8.1 body. Motion passed unanimously. 6.2 Identification of issues for further study 7. Connector Selection Tad Szostak - Connector Selection Criteria - Amateur friendly cabling system, economical tools, maintenance, in- stallation - Scalable from 100 Mb/s to many Gb/s - Easy daisy-chainable with splicing capability - ONE SOLUTION - Home wiring solution good for 10-20 years with transceiver upgrades - Direct fiber to PC buldheads from wall-plates to avoid copper EMI issues - Cabling system cost comparable (no more than 20% premium) to that of 100 ohm, UTP (CAT-5) Colin Whitby-Strevens - Connector requirements - Duplex assymetric, unique? - Suitable for equipment, dongles and wall-plates - low cost (very important) - Support speeds up to 4 GBaud - low loss (relative criterion) (1.5 dB for all connectors, 2 passive connections - Standardized or standardization in progress (EIA/TIA/ISO/etc.) - select, not invent - Simple installation, suitable for domestic applications - cable termination for wall plates, etc. - amateur installable with simple tools - Low to zero maintenance - OK to require pre-assembled patch cords - Suitable for MMF - suitable for POF may be desirable - support for eye safety in normal operation < 8. Future plans September in Nattick, Mass., on Tuesday the 23rd for next meeting. <<<<<<<< meeting adjourned at 4:00. >>>> =============================================================================== ACTION ITEMS 1. Jim Skidmore has an open action item to specify the speed-signal timings more precisely, and Mr. Brunker was asked to work with Jim on the issue. 2. Dave Johnson accepted the action item for communicating minimum risetime requirements for the p1394a speed signalling. 3. Colin Whitby-Strevens will initiate formal liaison/communication with TIA to add 50um fiber to their considerations, since we are addressing signal speeds up to 4 Gbps. Preferred method is via letter from P1394b chair. =============================================================================== ATTENDEES Richard Churchill 281-514-6984 richard.churchill@compaq.com Kazuki Nakamura +81 8275 3 5320 kazuki-n@dtinet.or.jp Mark Knecht 408-570-4097 mark_kencht@phoenix.com Ken Taylor 508-347-3309 ktaylor@bostonoptical.com Kazuyuki Abe 415-528-5904 kabe@sj-pceg.ccgw.nec.com Shuntaro Yamazaki +81 44 856 2082 yamazaki@ccm.cl.nec.co.jp Dan Rausch 408-435-6689 dan_rausch@hp.com Alistair Coles +44 117 922 8750 anc@hplb.hpl.hp.com Eric Hannah 408-765-4441 ehannah@mipos2.sc.intel.com James Kuo 408-721-7856 jamesk@galaxy.nsc.com Jim Busse 415-528-3810 jimb@ccgate.sj.nec.com Steve Bard 503-264-2923 steve_bard@ccm.jf.intel.com Ed Cady 503-359-4556 edcady@aol.com Karl Nakamura 408-433-4516 karln@lisl.com James Gay 512-891-2218 jimg@oakhill.sps.mot.com Hamid Shenavai No phone # given No email address given Farrokh Mottahedin 408-324-7934 fmottahe@qntm.com Reza Moattar 714-508-8731 reza.moattar@tsc.tdk.com Kenji Watanabe +81 3 5448 5362 nabeken@sslab.sony.co.jp Taka Fujimori +81 3 5488 6353 fujimori@arch.sony.co.jp Bill Northey 717-938-2119 northewa@bergelect.com Steve Midford 408-765-8370 steve_midford@ccm.sc.intel.com Peter Johansson 510-531-5472 pjohansson@aol.com Rahoul Puri 408-974-8598 rahoul@apple.com Kazuki Sogabe +81 6 466 5539 dw750134@jnet.sei.co.jp Tatsuya Arai 805-522-7958 tatsuyaa@hiroseusa.com Mike Brown 602-554-3713 mike_brown@ccm.ch.intel.com Jerry Hauck 408-765-5528 jerry_hauck@ccm.sc.intel.com David Wooten 281-518-7231 david.wooten@compaq.com Dave LaFollette 408-765-2587 dlafolle@mipos2.sc.intel.com Jack Merrow 425-486-2222 jmerrow@leviton-telecom.com Charles Brill 717-592-6198 cebrill@amp.com Peter Teng 408-588-5555 pteng@mail.com Claude Huss 650-237-2366 claude@mew.com Mike Wang 602-554-8555 mike_wang@ccm.ch.intel.com Dao-Long Chen 970-223-5100 x9461 dao-long.chen@symbios.com Michael Johas Teener 408-461-4901 mike@fireflyinc.com Susuma Morikura +81 6 906 4896 morikura@isl.mei.co.jp Takahiro Seki +81 462 30 6209 seki@lsis.crl.sony.co.jp P. Mohanraj 602-554-4243 palanisamy_mohanraj@ccm.ch.intel.com Paul Lind 602-554-3774 plind@sedona.intel.com Yukio Otobe +81 44 754 2532 otobe@flab.fujitsu.co.jp Farrukh Latif 610-712-7546 flatif@lucent.com Bijit Patel 610-712-2188 patelb@lucent.com Juan Pulido 512-984-5188 jmpulido@mmm.com Tad Szostak 512-984-3847 tszostak1@mmm.com Dave Brunker 630-527-2622 dbrunker@molex.com Max Bassler 630-527-4490 mbassler@molex.com David Johnson 972480-3632 dkjohnson@ti.com Dave Instone +44 1705 486363 dinstone@uk.xyratex.com IEEE P1394b Meeting August 8, 1997 Ala Moana Hotel, Honolulu, Hawaii Chairman David Wooten Secretary: Richard Churchill 1. Introductions, agenda and minutes of previous meeting 2. Liaison Reports 2.1 S100 2.2 S800 + combined 3. S800 connector 4. Coding 1. Introductions Introductions around, with comments by chair on the recent change in chairmanship, due to the resignation of Mike Teener. The agenda (abbreviated in anticipation of a lengthy discussion of item 4) was approved, with the understanding that additional topics will be dis- cussed following that item, as time permits. Minutes from the previous meeting were approved at a later point (see 8. below), when omission of this action was noted. 2. Liaison Reports 2.1 S100 Task Group, Taka Fujimori-san chair (Minutes of the 8/7/97 S100 Task Group are included by reference.) Taka discussed presentations regarding gradient index POF, and the draft of the S100 specification the group has been working on. This is to be voted on as a working document. The document was NOT accepted, but a re-vote will be held at a later date. Many abstentions due to lack of time to evaluate forced this delay. < PHY coding, PHY_DELAY, and Start-up procedure were discussed. < New action items included discussion of Root Contention. Next meeting is planned for September 2, in Tokyo, hosted by NEC. 2.2 S800 Long-haul Task Group, and Joint Long-haul session, Colin Whitby-Strevens chair (Minutes of the August 6, 1997, Joint Session, and the August 7, 1997, S800 Task Group are included by reference.) Colin discussed the relative time-scale of the S100 group as presented by Taka at the Joint session, which appears to be to complete the task prior to the full P1394b closure, and which may as a result entail non-1934b compatible features. S100 and S800 diverge, in that S100 envisions a full PHY, while the S800 group wants a sub-PHY. S100 plans to specify a wall-plate connector, while S800 does not. Decision make to recommend that the P1394b group select a single connector, device-to-dongle-to-wall-plate ... all the same to avoid confusion. S100 continues to work towards POF solutions, while the S800 group has selected a 50u glass fiber, with the direction of the P1394b group being a single medium. Power to wall-plate is still an open question. CEBus seems to solve the problem by use of a compound wall-plate, but isolated cable domains behind the wall-plate. Presentations from various sources were made. Discussion of coding schemes (4B/5B versus 8B/10B), with a listing of various plusses and minusses for each described. The S800 PMD meeting discussed signal integrity issues, with rise/fall time recommendations from seperate sources agreeing that a 500ps rise minimum is desired. Work on Speed signaling must continue, since Jim Skidmore was not present. S800-S3200 optical parameters were discussed. It is evident that S800 and S3200 over the same medium (a 50 nm MMF) is most likely accomplishable. A minor redefinition to say that the priority is S800, with interoperability with S1600 and S3200 implementations required, was made to the requirment. The GbE media specification was accepted as the model for the corresponding sections of the P1394b draft. This is linked to a tentative call for presentations on solutions to use this model for the appropriate parameters. Appropriate parameters for s800 through S3200 solutions should be presented. A request for a liaison letter from the P1394b chair to the TIA 41.8.1 and 41.8.2 committees to recommend consideration of 50nm MMF for their respective standards was moved and passed. A draft of the letter was presented for consideration, to be revised and voted upon later in this meeting. A preliminary list of connector requirements was presented, discussed and revised. (This was the result of presentation of requirements from two sources, which largely agreed.) A suggestion that a call for proposals be issued for selection of an existing connector (either already standardized or in the process of being standardized by an appropriate EIA/TIA/ISO/etc. bodies. < 3. S800 Connector Bill Northey, Berg Brief presentation of preliminary test results on a slightly modified connector was offered. Additional information should follow. < 4. Coding 4.1 How Many Coding Schemes? Wooten discussed the confusion of the diverging coding schemes for S100 and S800 implementations, and asked for a vote to re- confirm whether the committee of the whole still desires that we have a single coding for all P1394b implementations. Moved (Richard Churchill) and seconded (Steve Bard) that a single coding method be adopted for all 1394b environments. < < < < < Motion was called, with the comment that all who can raise a hand should do so. There were 35 yeas, 4 nays and 4 abstentions. It was observed that there were some who did not vote, even in abstention. Chair stated that the vote is of such importance that we will re-vote until everyone votes, even if only in abstention, and therefore recalled the question. Re-vote yielded 48 yeas, 5 nays and 7 abstentions, for a better than 9-to-1 margin over nays, and 4-to-1 over nays and abstentions combined. Those who voted against the motion were asked to comment. Taka observed that before we check the data, we may end up with an un- optimized solution. 4.2 Which Coding Scheme? Colin Whitby-Strevens presented material on 1394-1995 signaling, in preparation for a decision on Coding Scheme. *** hand-drawn diagram *** -1995 uses D/S differential bit level encoding for DATA signaling, with clock indication via the Strobe signal pair. " uses for arbitration signals that are unclocked differential handshakes (ones dominance) for simultaneous bi-direc- tional communication. " uses common mode signalling for speed signalling, 100 us long ... Coding scheme must support equivalent of the bi-directional signaling/arbitration mechanism of -1995 to be compatible with it (-1995). *** hand-drawn diagram *** 1394b will use only one signal pair for data, arbitration and speed signalling, such that (due to the cross of the pairs within a cable) transmit is only via TpB, and receiving only on TpA. The coding scheme we select must support such digital signaling for all three types of communication, and the belief is that both proposals support all these communications. The two scheme proposed are a 4B/5B scheme, and a modified 8B/10B scheme. - 4B/5B - FDDI 4B/5B encoding - NRZI encoding - Bit-level scrambling for all copper - not used for optical - MLT-3 for UTP only - not used for short haul copper and optical - Modified IBM 8B/10B - byte-level scrambling of data and control - IBM 8B/10B data encoding - special encoding for control symbols - more robust than IBM 8B/10B - all signalling binary - Common Features - encoding overhead the same for both (4-to-5) - Continuous operation at constant line rate (determined at start-up) - Control symbols for arbitration, speed signalling etc. - developed from existing solutions in the market - patents, believed to comply with IEEE patent policy - low complexity of digital logic - (analog is proposal and media dependent) < < [** The following is the text of Exhibit 1 of the licensing agree offered by IBM for the use of the IBM 8B/10B coding scheme in the 1394 standard. "Exhibit for IEEE 1394 standard "Exhibit 1 "This is an Exhibit to the Agreement Dated as of January 1, 1997 between IBM adn XXOTHER COMPANYXX. "I. 'Licensed Patents' shall mean: " Patent Group Number Date Issued ------------ ------ ----------- 1. U.S. Patent 4,486,739 December 4, 1984 Corrseponding Patents: --------------------- Canada 1190324 July 9, 1985 European Patent 97763 September 21, 1988 Germany 3378098 October 27, 1988 "In addition, Licensed Patents shall mean: "I. i) all other non-listed corresponding (or counterpart) patents to the foregoing patents, whether issued or issuing in the future; and ii) all patents which are reissues, divisions, continuations, or extensions of any of the foregoing patents. "II. 'Licensed Products' shall mean products which: i) are primarily designed to convert data from an eight (8) -bit code to a ten (10 -bit code or from a ten (10) -bit code to an eight (8) -bit code; and ii) conform to and operate in compliance with the specifications of a Supported Standard. "III. 'Supported Standard' shall mean the IEEE 1394 Standard promulgated by the Institute of Electrical and Electronics Engineers (IEEE). "IV. License Fee - $5,000" The text of the Hyundai letter as subsequently presented to the group secretary is as follows. "HYUNDAI Hyundai Electronics Industries Co., Ltd. -------------------- Hyundai Jeonja Bldg. 66 Chuckseon-dong, Chongro-ku Seoul, Korea Tlx:k29793/4 HDETN Tel:398-4500 Fax:744-2145/7 "June 11, 1997 "To Whom It May Concern (DAVIC1.2) "This is to advise you that Hyundai Electronics Industries Co., Ltd. is willing to grant to any requesting party a nonexclusive license under the claims of Hyundai's U.S. Patent No. 4,530,088 entitled 'Group Coding System for Serial Data Transmission' if infringement of this patent is necessary to comply with DAVIC1.2, when such standard is issued as an Davic Stardard. "Each grant of license will be under a separate agreement and will be effective upon a payment to Hyundai of an one-time fee of ten thousand U.S. dollars (U.S. $10,000). In addition, such agreement will require each requestor to grant an option to Hyundai to acquire a license under requestor's patents for the same on-time fee. "Upon the adoption of the DAVIC1.2 as a standard, parties who wish a license should contact the undersigned at the above address or our U.S. counsel at the address identified below. "Very truly yours, [signature] D.S. Chung Director Patent Department "U.S. Counsel ------------- Robert C. Colwell, Esq. TOWNSEND and TOWNSEND and CREW 379 Lytton Avenue Palo Alto, California 94301-1431 Tel: 415-326-2400 Fax: 415-326-2422" Those wishing a copy of either may contact the group secretary, Richard Churchill, 'richard.churchill@compaq.com'. Readers are reminded that such agreements are entirely under the control and at the discretion of entities holding the patents sited, and the secretary does not warrant any terms stated or implied in this exhibit. **] - Encoding Schemes - 1 *** table *** Advantages Disadvantages ----------- ----------------------- ------------------------------ 4B/5B 2 errors in 40ns for an Unbounded disparity erroneous GRANT Short code allows fast serial/parallel conv. (40ns? less latency) Based on FDDI (in market at 100 Mbps) 8B/10B Based on IBM 8B/10B, 2 errors in 80ns for an (in market at multi- erroneous grant gigabit rates) Longer latency equivalent Detects three single to approx 7.5m line error events in a length packet (meets IEEE standard) Bounded disparity (+/- 6) Hamming distance of 2 between control and data symbols, as well as between the control codes Spare control codes - S100 UTP advantages/disadvantages Advantages Disadvantages ----------- ----------------------- ----------------------------- 4B/5B 100m at equivalent transmit ... - S100 optical advantages/disadvantages Advantages Disadvantages ----------- ----------------------- ----------------------------- - S800 short-haul copper advantages/disadvantages Advantages Disadvantages ----------- ----------------------- ----------------------------- - S800 optical advantages/disadvantages Advantages Disadvantages ----------- ----------------------- ----------------------------- Moved (Colin Whitby-Strevens) and seconded (Richard Churchill) that the p1394b study group adopt the modified 8B/10B coding scheme proposal as the basis for its specification. Taka stated that he wishes to hear presentations, and pointed out that Sony has working prototypes of the 4B/5B scheme. Comments to the effect that some of the objections to 8B/10B do not necessarily apply ... 4-bit boundaries are problematic. Alistair stated that a three-level coding scheme that is DC balanced can be based upon the modified 8B/10B scheme ... Number of errors required to turn a signal into an erroneous grant can be corrected by reassignment to obtain a Hamming distance of 6 between grant and other symbols. Chair asked that we focus on the native attributes, and not upon features that several of the engineers present could correct with some effort. Colin observed that a single bit error can convert a 4B/5B symbol into a different valid symbol. Taka replied that 4B/5B repetition of GRANT helps with this. Comment that IBM 8B/10B is very robust due to the coding mechanism and checking. Only 269 of 1024 available symbols are used, and running disparity is monitored, so that single bit errors will be caught within 16 bits. Observation made that an unbalanced scheme is painful in implement- ation over optical links ... Question raised regarding the objectives ... 4B/5B was chosen by FDDI for expediency ... If had to do over, would likely pick 8B/10B. Several expressed view that 8B/10B has been broadly adopted due to its cost advantages. The question was called by Steve Bard. The vote was 46 yeas, 8 nays, and 5 abstentions, and the motion was carried. As the motion carried by a substantial margin, the chair declared that a super-majority of 75% will be required to overturn this decision. Taka (who opposed) expressed concern that S100 may not be opti- mized due to this decision. <<<<<<<< Recess for lunch at 11:55, resumed at 1:10 >>>> 5. Presentation by .... Regarding the "EBU/SMPTE Task Force for the Harmonization of Standards for the Exchange of Program Material As Bit Streams" Request For Technology ... [** With apologies, the secretary notes that his record of this presentation was lost when someone kicked the power cord powering his system, thus turning it off. **] Information regarding the RFT, and about the standards groups involved, may be obtained from the following web sites: www.smpte.org www.ebu.ch The following files are available regarding this request: "TFHSRPT1.PDF" "RFT.DOC" Moved (R. Churchill) and seconded (S. Bard) that Colin Whitby- Strevens be authorized to act on our groups behalf to establish contact with this organization, and represent us before it. Motion passed unanimously. 6. Connector requirements Colin Whitby-Strevens - Connector requirements - 1 - Cost of ownership, etc. - Low cost (v important) - enables system (cable, connectors, any other necessary hardware installation cost comparable to UTP-5) - Good for home wiring installation with 20 year lifetime with transceiver upgrades for higher bandwidth - Simple installation, suitable for domestic application - simple cable termination for wall plates, etc. - amateur installation, simple, economical tools - low to zero maintenance - amateur maintenance - no problem from dust, etc. - Standardized or standardization in progress (EIA/TIA, ISO, IEC, etc. ... - select, not invent - Connector requirements - 2 - Function - Duplex, asymmetric, differentiated (keying, etc.) - Support for speeds from 125 MBaud up to 4 GBaud - Suitable for equipment, dongles, and wall-plates - allow any to any connection with patch cords - OK to require pre-assembled patch cords - Small size (e.g. RJ-45 aperature compatible) - low loss - 1.5dB link connector budget - at least 2 passive connections, support for additional splices desirable - Format suitable for MMF - suitable for pof may be desirable - support for eye-safety in normal use - Assist physical safety Proposals are called for, to be reviewed in the PMD task force. We will need to know patent issues, but this is subsumed into the "standard" requirement. A draft letter of a contact/request letter, to be sent by the chair of the p1394b task group, second to the motion not being needed as this is a request from a sub-group, presented and moved by Colin. A move to table was offered and withdrawn. Amendment to refer the question back to the committee offered and accepted as friendly. Amended motion passed unanimously. Contact with ISO/IEC SC 25 chair: Professor D. Popovic University of Bremen Bruchwetern 6B 28 Bremen 33 Germany fax: 49-421-272-298 or with SC 25 secretary/WG3 convenor: Dr. Walter van Pattay Secretary ISO/IEC JTC1/SC25 Siemens AG Techincal Regulations adn Standardization ZT TNI Otto-Hahn-Ring 6 D-81730 Munchen Germany fax: 49-89-636-43-891 walter.pattay@mchp.siemens.de was suggested/requested. 7. Mike Fogg (AMP) Connector presentation - Differences between 100 Mbps and .... *** graphs [3] *** - P1394a, Differential Impedance Profile Between A3 and B3 *** graph *** - P1394a, Simulated Differential Impedance ... *** graph *** *** graphs [4] *** *** graphs [4] *** - Output Eye pattern, 4m, 30 AWG, 3.5 GHz *** graph *** - Output eye pattern, 10m, 24 AWG, 3.5 GHz - 4. EMI/RFI Characteristics Shielding Effectiveness Information Is EMI controlled at Board Level, or at Bulkhead (or Both)? Significant spectral content at higher frequencies. Asymmetrical performance creates common mode noise Cable assembly acts more like a coaxial assembly Seams, aperatures and other breaks in shield will radiate. System noise coupled onto signal lines creates common mode noise Source induced skew or cable induced skew creates common node noise Many Cables .... ... 8. Minutes of last meeting Minutes from last meeting accepted unanimously without revision. 9. PAR Status, call for patents, and Editor's report 9.1 PAR Status, David Wooten We have had editorial changes requested on the last meeting, but no action on the PAR by the MPSC (?). 9.2 Call for patents, David Wooten If you know of any patents you hold that may be contained within the p1394b standard, or that may be included in that standard, you are called to present such information. Call applies to any and all parties ... 9.3 Editor's report, Eric Hannah Eric Hannah offered changes, etc., to the spec as last presented. The launch voltage was altered to reduce to sub volt levels for the differential signal. Concensus that the max should be raised to 800 mV noted and to be incorporated. Editorial change to make the nomenclature both clearer and more consistent with -1995 suggested and accepted. Assorted other editorial changes suggested, some accepted and some rejected. The validity and desirability of various minimum times was discussed, with general agreement that the minimum rise/fall times should be the same, even though the max times are derived from theory. Receiver characteristics received a comparable beating, but with less change (cf table 5-2). Nominal numbers need to be correct, and support for allowed ranges stated. Table 5-3 was also critiqued. Questions were raised regarding the note stating number of connection ports supported ... Clause 7 is currently based upon modified 8B/10B, which as of this meeting is the accepted base ... Alistair asked to offer any needed corrections. Clause 8 -- speed codes should be the same length in time, regard- less of the speed at which the code is transmitted. Teener will provide a table to replace the present table 8-1. Clause 9 -- Ports -- Suggestion made by Teener that an open issues section be placed at the start of each section/clause, to allow readers to more easily identify areas of concern, or where some further contribution is needed. Observation that the changes to Suspend/Resume for P1394a will entail some changes to the planned detection schemes of P1394b was made. 10. Start-up Alistair Coles, HP Labs [** Readers are referred to the digital copy of this presentation available on the group ftp site. **] - Start-up with modified 8B/10B Scheme - Background - During Beta-mode Start-up, handshake signals are exchanged between 1394b PHYs so that bit, codeword and scrambler synchronization can occur - In June, the following handshake patterns were proposed: s1(ESL) = [k28.5, D10.2] = 0011111010 0101010101 s2(ESR) = [k28.5, D21.5] = 0011111010 1010101010 - These ... - Concerns with June proposal: - S1, S2 and S3 ... - New Proposal - S1(ESL) = REQUEST/GRANT (i.e. stream of Cx's derived from scrambled control state 00001 - S2(ESR) = IDLE (i.e. stram of Cx's derived from scrambled control state 00000). !!! Bit, codeword and scrambler synchronization can be achieved during S1 and S2 !!! CODEWORD SYNCHRONIZATION - Scrambler (1) *** diagram *** - Scrambler (2) *** diagram *** - Scrambler output during start-up *** diagram *** - Descrambler Synchronization - During start up (and at any other time IDLE is received), decoded control values contain regular samples of trans- mitter scrambler state. (2 samples per received codeword). - These samples can be used to synchronize the descrambler, using extremely simple hardware. - 11 samples of the scrambler state are required to guarantee synchronization of descramber, i.e. synchronization achieved in 6 codewords deration. - This is an example of a distributed sample scrambler. (See: "Synchronization of shift register generators in distributed sample scramblers," Kim, S.C., and Lee, B.G., IEEE Trans. Comms., Vol.42, Feb. 1994, pp. 1400-1408.) - Descrambler *** diagram *** - Speed Negotiation - For speed negotiation, each node issues its speed caps by sending a repeating sequence of control states of the form: ... ESCAPE SPEED+ SPEED+ SPEED+ ESCAPE SPEED+ ... The number of SPEED+ states following each escape indicates the maximum speed cap. of the node, in a similar way to the number of speed states following data prefix indicates the rate of following packet. *** TABLE *** - Each node compares received max. speed indication with its own speed, and changes speed to the lower of the two (unless this equals the current speed). - ESCAPE defined as control state 1100 (currently spare). - Start-up sequence *** diagram *** - Possible enhancement to start up - IBM 8B/10B code is implemented for Fibre Channel and GbE uses K28.5 character for codeword synch. To maintain this, insert K28.5 characters into S1 and S2 control codeword scheme (e.g. one K28.5 character in every 16 Cx characters. - This would also allow S1 and S2 to be distinguished from IDLE and REQUEST during normal arbitration. -> Potential for one node to request start-up. - Root Contention - Root contention can be resolved prospectively during start-up by nodes negotiating in advance their behaviour in the event of a contention. Negotiation not needed after each Bus Reset. - Concurrent with speed negotiation, nodes could exchange random "bids" to establish wich will become root in the event of contention. e.g. bid for "I will become root" = ESCAPE, CHILD_NOTIFY bid for "I will become chile" = ESCAPE, PARENT_NOTIFY - Bids are insered in max. speed indication stream. After issuing first bid, each node waits to receive a bid. - if the two bids are different, contention is resolved. - if the bids are the same, node issues a new bid. - Round trip delay can be eliminated by nodes issuing a burst of say 8 bids, and comparing these sequentially with received bids. First pair of differing bids resolves contention. - Root Contention example *** diagram *** *** table *** - 1/256 chance of no result after 8 bids: if so, each node issues new bids, and process repeats. - Root Contention example (cont.) *** diagram *** - Summary - During start-up proposed patterns are: S1(ESL) = REQUEST = control state 0001 S2(ESR) = IDLE = control state 0000 Easy codeword and scrambler synchronization. - At end of S2 (ESR), transmit max. speed capability in form: ESCAPE SPEED+ ... ESCAPE SPEED+ ... Number of SPEED+ states indicates max. speed caps. - Alternative Root Contention resolution method described using prospective bidding during start-up. Note: Other recent proposals to solve long distance root contention also work with modified 8B/10B coding. 11. Root contention David LaFollette, Intel [** Readers are referred to the digital copy of this presentation available on the group ftp site. **] - - Contention Times - Alternative Proposals - - Avoiding SubPHY Becoming Root - The subPHY could simply forward parent and child notify singal - The root contention constant would have to be really big - big enough to allow for 3 cables and 2 intermediate ... - SubPHY Tree-ID *** state diagram *** < - Example of Contention *** diagram *** - Example of NO Contention *** diagram *** 12. Decide if AC or DC coupled No discussion beyond "We need to answer this," was noted. 13. Which columns do we fill in? What are the open issues we have address? < < Open Issues ------------------------------------------------------------------ PHY-Link interface for S800+ - requirement Suspend/Resume over long-haul Beta-mode start up -- S100 or S800? Is this a property of SubPHYs? What can a PHY drive directly? How slow does a Beta-mode PHY have to be able to run? Cable IDs (slow/fast)? < Properties of SubPHYs < Colin Whitby-Strevens - Properties of the sub-PHY - Two port PHY repeater function - one port is 1394b Cu and one long-haul - DS-capble on CU side - Beta-only on optic side - may be beta capable on Cu side (necessarily capable for S800+) - At least one node ID in a mated pair (two ends of same cable) - Does NOT have a Link (its delay is intrinsic to the cable within which it is embedded) (can attach long-haul to a regular PHY) - Must be able to speed filter in both directions (speed filtering rules apply in both directions) - Opportunity to use gap reduction tricks in beta mode (performance improvements) - Must retime signals due to prospect of D/S to Beta transitions - Needs some form of elasticity buffer - Should refrain from being root < < < < < 14. Future meetings September 23 and 24, in Natick, Mass. October, week of 13-17, with the TA meeting in Phoenix November ?? and ?? in Houston?, or possibly with P1394a (Dec. 3-5) December 8 and 9, in Orlando. Difficulties in agreement to a November meeting lead to the idea of NOT having the meeting with the TA in October, but later, so as to more evenly space three meetings over the four months. 15. Action items - Various to respond to call for connector proposals. - Bill Northey to provide additional test results on connector for S800 and beyond. - Colin Whitby-Strevens to develop contact with IBU/SMPTE. - Teener to provide table for speed signalling. - Wooten to determine possible alternative meeting sites for a late October meeting in place of co-location with TA, and to clarify meeting schedule. - Steve Bard to examine Beta mode suspend/resume, and detection in general. - Problem of deciding whether we need to go to 2 node IDs per cable or should stay with one per cable is set as an action item for the committee of the whole. - Action items not completed for this meeting are carried over. 17. Adjournment =============================================================================== Attendence List Richard Churchill 281-514-6984 richard.churchill@compaq.com Mark Knecht 408-570-1097 mark_knecht@phoenix.com Dan Paley 408-570-1256 dan_paley@phoenix.com Brad Saunders 714-221-6513 bradley.saunders@rss.rockwell.com David Wooten 281-518-7231 david.wooten@compaq.com Steve Bard 503-264-2923 steve_bard@ccm.jf.intel.com Alistair Coles +44 117 922 8750 anc@hplb.hpl.hp.com Dan Rausch 408-435-6689 dan_rausch@hp.com Colin Whitby-Strevens +44 1454 611500 colinws@bristol.st.com Taka Fujimori +81 5488 6353 fujimori@cv.sony.co.jp Dao-Long Chen 970-223-5100 x9461 dao-long.chen@symbios.com Kazuyuki Abe 415-528-5904 kabe@sj-pceg.ccgw.nec.com Paul Lind 602-554-3774 plind@sedona.intel.com P. Mohanraj 602-554-4243 paknisamy_mohanraj@ccm.ch.intel.com Jim Busse 415-528-3810 jimb@ccgate.sj.nec.com Gene Milligan 405-324-3070 gene_milligan@notes.segate.com Daniel Meirsman +32 16 390 733 daniel.meirsman@leu.ce.philips.com Geert Knapen +32 16 390 734 geert.knapen@innet.be Takahiro Seki +81 462 30 6209 seki@lsis.crl.sony.co.jp Dave LaFollette 408-765-2587 dlafolle@mipos2.sc.intel.com Jerry Hauck 408-765-5528 jerry_hauck@ccm.sc.intel.com Mike Wang 602=554-8555 mike_wang@ccm.ch.intel.com Jim Skidmore 972-480-2094 j_skidmore@ti.com David Johnson 972-480-3632 dkjohnson@ti.com Max Bassler 630-527-4490 mbassler@molex.com Dave Brunker 630-527-2622 dbrunker@molex.com James Gay 512-891-2218 jimg@oakhill.sps.mot.com Ed Butler 602-554-0751 ebuter@sedona.intel.com Jim Doyle 602-554-2051 jdoyle@sedona.intel.com Dave Instone +44 1705 486363 dinstone@uk.xyratex.com Johathan Thatcher 507-253-2867 jonathan_thatcher@vnet.ibm.com Bill Prouty 916-785-4631 billp@hprnd.rose.hp.com Bill Pherigo 970-229-3586 wlp@fc.hp.com Juan Pulido 512-984-5188 jmpulido@mmm.com Roy Yasoshima 415-858-1000 yaso@masca.com Doug Hagerman 508-841-2145 hagerman@mail.dec.com Michael Nguyen 408-894-3542 michael.nguyen@fcpa.fujitsu.com Steven Midford 408-765-8370 steve_midford@ccm.sc.intel.com Bill Herb 717-810-3892 bill.herb@amp.com Mike Fogg 717-986-5802 mike.fogg@amp.com Hamid Shenavai 408-922-9614 hshenava@fmi.fujitsu.com Farrokh Mottahedin 408-324-7934 fmottahe@qntm.com Reza Moattar 714-508-8731 reza.moattar@tsc.tdk.com David Deming 408-338-9285 ddeming@scruznet.com Hirokazu Mamezaki +81 22 347 1128 mamezaki@ffm.fujifilm.co.jp Yasumasa Hasegawa +81 22 347 1128 yasumasa@ffm.fujifilm.co.jp Farrukh Latif 610-712-7546 flatif@lucent.com Paul S. Levy 602-752-6382 paul.levy@vlsi.com Zack Ciccone 602-752-6233 zack.ciccone@tempe.vlsi.com Tatsuya Arai 805-522-7958 tatsuyaa@hiroseusa.com Michael Johas Teener 408-461-4901 mike@fireflyinc.com Kenji Watanabe +81 3 5448 5362 mabeken@sslab.sony.co.jp Dick Scheel 408-955-4295 dicks@lsi.sel.sony.com Bijit Patel 610-712-2188 patelb@lucent.com Shuntaro Yamazaki +81 44 856 2082 yamazaki@ccm.cl.nec.co.jp Nelson Arata 408-721-4979 nelson.arata@nsc.com Peter Teng 408-588-5555 pteng@mail.com Phil Young +44 1908 837247 youngp@euk.nec.co.uk James Kuo 408-721-7856 jamesk@galaxy.nsc.com Eric Hannah 408-765-4441 ehannah@mipos2.sc.intel.com Jack Serafin 602-554-0459 jack_a_serafin@ccm.ch.intel.com Kazuki Nakamura +81 8275 3 5320 kazuki-n@dtinet.or.jp Karl Nakamura 408-433-4516 karln@lsil.com Rahoul Puri 408-974-8598 rahoul@apple.com --Cyberdog-MixedBoundary-000E88B8--