X3T9.2/92-72 Date: March 17, 1992 To: X3T9.2 Membership From: Lawrence J. Lamers, X3T9.2 Secretary Bill Spence, Chair SPI Working Group Subject: March 17, 1992 SPI Working Group Meeting Minutes Bill Spence called the meeting to order at 1:00 p.m. March 17, 1992. He thanked Marilyn Martinez of AMCC for hosting the meeting. The SPI working group is the umbrella for all contact, connector, cable, transceiver issues related to SCSI-3 Parallel Interface (SPI). Bill Spence chairs this working group which is chartered with developing a set of recommendations for the SPI standard that will improve the physical plant to a more reliable state. The final agenda was as follows: 1. Review of 3/13/92 SPI Working Draft Document--X3T9.2/91-10R3--Kurt Chan 2. Review of Fast Timing Skew Budget--X3T9.2/92-62--Kurt Chan 3. Node Capacitance Revisited--Bill Ham 4. Driver Transition Times 5. Receiver Filters 6. Impedance and other Cable Specifications--X3T9.2/92-57--from Tom Debiec 7. Mixed Width RESERVED and TERMPWR Issues--X3T9.2/92-50--Bill Spence The following people attended the meeting: Name Status Organization Phone Number ------------------------- ------ ------------------------------ -------------- Mr. Thomas Newman P Adaptec, Inc. (408) 945-8600 Mr. Al Wilhelm A Adaptec, Inc. (408) 945-2525 Mr. Sassan Teymouri A Advanced Micro Devices (408) 235-8616 Mr. Alan Folmsbee V Advanced Micro Devices (408) 235-7257 Mr. Tony Castillo O Aeronics, Inc. (512) 258-2303 Mr. Russell D. Moser V AMP, Inc. (717) 986-5022 Mr. Scott Smyers P Apple Computer (408) 974-7057 Mr. John Geldman A Cirrus Logic Inc. (510) 226-2368 Mr. Charles Monia P Digital Equipment Corp. (508) 841-6757 Dr. William Ham A Digital Equipment Corp. (508) 841-2629 Mr. Rick Muething S Emulex (Consultant) (510) 248-0787 Mr. Skip Jones A Emulex Corp. (714) 668-5058 Mr. D. W. Spence A ENDL Associates (512) 255-0339 Mr. Denni Rash O FileNet Corp. (714) 966-3209 Mr. Robert Liu P Fujitsu Computer Products,Am (408) 894-3790 Mr. Jeffrey L. Williams P Hewlett Packard Co. (208) 323-5030 Mr. Kurt Chan A Hewlett Packard Co. (916) 785-5621 Mr. Howard Wang O Hitachi Computer Products (408) 986-9770 x207 Mr. George Penokie P IBM Corp. (507) 253-5208 Mr. William D'Andrea S J.S.T. Corporation (714) 669-8035 Mr. Ronald Diotallevi V JST Corp. (708) 803-3300 Mr. Lawrence J. Lamers P Maxtor Corp. (408) 432-3889 Mr. Raymond C. Yule O Micropolis Corp. (818) 718-7803 Mr. David Steele S NCR Corp. (719) 596-5795 Mr. Gene Milligan A Seagate Technology (405) 324-3070 Mr. Gerald Houlder A Seagate Technology (612) 844-5869 Mr. Alan Wetzel O Texas Instruments (214) 997-2370 Mr. Bob Mammano V Unitrode Integrated Circuits (603) 429-8628 Mr. Tak Asami O Western Digital (714) 932-7621 29 People Present Status Key: P Principal A Alternate O Observer S Special Interest (frequent visitor) V Visitor Results of Meeting 1. Review of SPI Working Draft Document Kurt pointed out the revisions since previous distribution, which mainly just documented the results of February's SPI meeting. 2. Review of Fast Timing Skew Budget The current specification of skew budget at the connector is not reasonable in either SE or DIFF case. Kurt has developed a skew budget using published data showing skew at the connector and skew at the chip. Working with published data workable SE implementation is possible. However, DIFF is a problem, based on the published data. A proposed compromise skew budget will be developed based on differential that will work for SE, but be tighter than current numbers. Kurt presented his budget. SCSI-2 had a 31ns difference between transmit and receive skew. The proposed SCSI-3 is 21ns. (See X3T9.2/92-62). Once silicon numbers established extrapolate to connector. The differential driver mfg's need to buy into the 6ns number. 3. Node Capacitance Revisited (Ham) Bill Ham reviewed the considerations affected by our tentative decision to reduce the node capacitance limit to 20 pF (from 25). He reports that implementations that run fast successfully can be over 20 pF. Many existing devices meet the 25 pF in SCSI-2, however a number of chips do not even get below 20 pF. Input filters can affect the measurement of capacitance. How to measure capacitance is also important. Active negation (2 transistors on a pad) increase capacitance significantly. Things affecting pF: active negation & double pads on chips (pF/pad) (18), traces and stub length (2/in), switchable terminators (3-x) , and input RC filters, connectors (2-3). The signal sees each lumped capacitance as an individual challenge; it does not see the total bus capacitance at once. It was agreed that holding even 25 pF node capacitance is a challenge; 20 pF is not practical. The wisdom of our forefathers has shown through and we are now back to the historical precedent of 25 pF. Some of the discussion in this item slopped over into the question of signal transition time, because it was stated that more node and other lumped capacitance is tolerable with longer transition times. The results of this discussion are reported in the next item. The test circuit to check the rise time will have some lumped capacitance. Steele: 47 ohms to 2.5 VDC 20 pF measured 10-90 modeled real world silicon very well. This due for an annex. The 20 pF represents transmission line load. 4. Driver Transition Times This discussion raised again the question of what is being specified. It was reiterated that it is the 10-90% transition time measured at a chip terminal when it is in a specified test circuit. This circuit was presented in David Steele's 91-64R3 and is to be included in the SPI document. David proposed, however, that the parameters be trimmed to 47 ohms, 20 pF, and 2.5 vdc. With that out of the way, the question of what the minimum transition time should be was then faced. There had been considerable agreement at earlier meetings on a figure of 5 ns, but the silicon attendance was limited and there were reservations. This time, there was good silicon attendance and those with reservations agreed to withdraw them. So 5 ns is in. 5. Receiver Filters One approach to this subject was to utilize the signal space diagram which Florin Oprescu had presented in February. Am optimum filter would reject all signals contained completely within Florin's signal space. One problem is that the diagram Florin presented was so restrictive that it could not be applied to equipment intended to operate faster than about 2.5 megatransfers per second. It was anticipated that Florin might bring in diagrams for faster transfers, but that did not happen. Bill Spence opined that Florin's work could be utilized simply by reducing the time axis calibration by a factor of 4 (to convert from 2.5 to 10 megatransfers). It would then supposedly describe the most restrictive filter which could be employed and still permit fast SCSI operation. The question then was faced as to whether this approach was more or less helpful than the simple signal transition type test which David Steele had proposed in 91-64R3. An advantage of the latter approach is that it permits a much simpler test for conformance. Avoiding double clocking is a significant challenge; it represents the single most important data integrity issue on the SCSI bus. Time domain filters can help. AMD feels that the filter will be different for different transfer speeds. AMD has been granted a patent and will abide by the ANSI patent policy. Sassan Teymouri showed a glitch scenario that the AMD filter could handle. Specifically the glitch that could occur due to cross talk when switching the data lines prior to asserting REQ. Ref Kurt's paper from years past, that stated this glitch was 10-15 ns wide. Tak Asami of WD showed a SCSI ACK NET WITH WIRED-OR DRIVER that illustrated glitches on a transition. The active negation driver pushed the glitch caused during negation up into a region that is a don't care. The assertion glitch is a function of cable length, stub length, and slew rate (X3T9,2/92- 61). Sasson suggested that a 2 volt spike be eaten. John Geldman stated that we need to concentrate on the transition areas. Need to center on the double clocking problem. Bill Spence stated that we need some new words to shoot at and John agreed to bring in some. 6. Impedance and other Cable Specifications [Debiec] Bill Spence presented Tom Debiec's proposal for cable specs. The values will be translated into per meter numbers. PVC should not be used as a dielectric. Bill Spence agreed to integrate Tom's material into the existing standard. Larry Lamers asked that an equivalent flat cable spec is needed for in-the-box connections both high-density and low-density that matches up with the external cables spec'd. Bill Spence opined that the new impedance specs we have come up with provide adequate homogeneity in the SCSI bus and are achievable in all cable types with proper choices of wire gauge and dielectric. 7. Mixed Width RESERVED & TERMPWR Issues (Spence) Bills recent implementation of a 16-bit cable with an 8-bit device in the middle. Folks not following the letter of the standard get in trouble when the reserved lines are grounded in the device. Annex A of SPI illustrates a solution. Recommend deleting the ground option, RESERVED lines should be open or terminated. Doesn't work cause if terminated and TERMPWR applied the regulator will turn off. It seems like some real consideration should be given to having RESERVED lines be not connected in all devices. The alternative seems to be that otherwise our RESERVED lines are actually not available. 8. Glitch Free Power Up/Down Bill Ham raised an issue of whether glitch free operation will be spec'd in SCSI-3. The driver shall not address the bus during power up or power down as long as the power transition occurs in less than 5 milliseconds to stabilized power. Bill Ham will bring in a proposal. 9. Meeting Schedule It was semi-agreed that the silicon people would try all to be present at the April SPI meeting in St. Petersburg Beach to finish up the receiver filter matter. Then the May SPI meeting would concentrate on finalizing all cable issues. The SPI Working Group meeting for April is on Tuesday, April 28, 1992 at 9:00 am at the Tradewinds Hotel in St. Petersburg, FL. The SPI Working Group meeting for May is on Tuesday, May 19, 1992 at 1:00 pm at the Holiday Inn in Harrisburg, PA. The meeting adjourned at 6:30 p.m.