Minutes of SPI-2 Working Group - April 22-23, 1996

Lohmeyer, John JLOHMEYE at cossymwest.co.symbios.com
Tue May 14 11:05:00 PDT 1996


* From the SCSI Reflector, posted by:
* "Lohmeyer, John" <JLOHMEYE at COSSYMWEST.CO.SYMBIOS.COM>
*

Minutes of SPI-2 Working Group                   X3T10/96-157r0



Accredited Standards Committee*
X3, Information Technology
                                              Doc. No.: X3T10/96-157r0
                                                  Date: May 14, 1996
                                               Project: 1142-D
                                             Ref. Doc.:
                                              Reply to: John Lohmeyer

To:       Membership of X3T10

From:     Bill Ham, SPI-2 Technical Editor
          Larry Lamers, Vice Chair X3T10
          John Lohmeyer, Chair X3T10

Subject:  Minutes of SPI-2 Working Group
          April 22-23, 1996 -- Milpitas, CA


                            Agenda

1. Opening Remarks

2. Approval of Agenda

3. Attendance and Membership

4. Test Data (96-141) [Ham]

5. LVD SCSI Driver Specification Presentation (96-145r0) [Moore]

6. Standing Waves [Bridgewater]

7. Glitch Issues
  7.1 Releasing Bus from Active Negation [Uber]
  7.2  Digital  Filtering of  Negation Release Glitches  (96-158r0)
     [Lohmeyer]
  7.3 LVD Release Glitches Proposal (96-160) [Kakirian]
  7.4 Glitch Filtering [Ham]

8. Transmission Line Model (96-123 & -124) [Gingerich]

9. Timing Budget [Ham]

10. Hot Plugging [Ham]

11. SPI-2 Document Review (X3T10/1142D) [Ham]

12. Summary of Meeting Results

13. Meeting Schedule

14. Adjournment




                          Results of Meeting

1.   Opening Remarks

John Lohmeyer, the X3T10 Chair, called the meeting to order at 9:00
a.m., Monday April 22, 1996.  He thanked Norm Harris of Adaptec for
hosting the meeting.

As is customary, the people attending introduced themselves and a copy
of the attendance list was circulated.

2.   Approval of Agenda

The agenda was approved with the following additions:

     Asymmetrical driver specifications (actually covered under agenda
     item 5)
     Hot Plugging

3.   Attendance and Membership

Attendance at working group meetings does not count toward minimum
attendance requirements for X3T10 membership.  Working group meetings
are open to any person or organization directly and materially
affected by X3T10's scope of work.  The following people attended the
meeting:

         Name          S        Organization         Electronic Mail   
Address
 ---------------------- -- -------------------------   
 -------------------------
Mr. Norm Harris        P  Adaptec, Inc.   
            nharris at eng.adaptec.com
Mr. Lawrence J. Lamers A  Adaptec, Inc.             ljlamers at aol.com
Mr. Tak Asami          A# Adaptec, Inc.             asami at itc.adaptec.com
Mr. Wally Bridgewater  V  Adaptec, Inc.             wally at eng.adaptec.com
Mr. Richard Moore      V  Adaptec, Inc.   
            richard_moore at corp.adaptec
                                                    .com
Mr. Tom Schneider      V  Adaptec, Inc.   
            schneid at itc.adaptec.com
Mr. Chris Burns        V  Adaptec, Inc.   
            chrisb at eng.adaptec.com
Mr. Shahe Krakirian    V  Adaptec, Inc.   
            shahek at eng.adaptec.com
Mr. Louis Grantham     P  Dallas Semiconductor      grantham at dalsemi.com
Mr. Siegfried Schmalz  V  Dallas Semiconductor      schmalz at dalsemi.com
Dr. William Ham        A# Digital Equipment Corp.   
  ham at subsys.enet.dec.com
Mr. Richard Greenberg  V  IBM Corp.                 richg at vnet.ibm.com
Mr. Dean Wallace       P  Linfinity Micro   
          75671.3443 at compuserve.com
Mr. Wayne E. Werner    O  Lucent Technologies       wew at aluxpo.lucent.com
Mr. Ting Li Chan       A  QLogic Corp.              t_chan at qlc.com
Mr. Richard Uber       V  Quantum Corp.             duber at tdh.qntm.com
Mr. Gene Milligan      P  Seagate Technology        Gene_Milligan at notes.
                                                    seagate.com
Mr. James Whitworth    A# Seagate Technology   
       james.whitworth at conner.com
Mr. Brian N. Davis     A# Seagate Technology   
       brian_davis at notes.seagate
                                                    .com
Mr. William C. Gintz   V  Seagate Technology        bill.gintz at conner.com
Mr. Dave Guss          P  Silicon Systems, Inc.   
    dave.guss at tus.ssi1.com
Mr. Vit Novak          A  Sun Microsystems, Inc.    vit.novak at sun.com
Mr. John Lohmeyer      P  Symbios Logic Inc.   
       john.lohmeyer at symbios.com
Mr. Frank Gasparik     V  Symbios Logic Inc.   
       frank.gasparik at symbios.com
Mr. Pete Tobias        A  Tandem Computers   
         tobias_pete at tandem.com
Mr. Matt Kaltenbach    V  The Panda Project         mattk1 at msn.com
Mr. Paul D. Aloisi     P  Unitrode Integrated       Aloisi at uicc.com
                          Circuits

27 People Present

Status Key:  P    -  Principal
             A,A# -  Alternate
             O    -  Observer
             L    -  Liaison
             V    -  Visitor

4.   Test Data (96-141) [Ham]

Bill Ham presented data that indicated a 15 meter LVD segment with 15
20pf/20pf/10pf loads on 8-inch centers is possible.  A possible
alternative is 10pf/10pf/5pf loads on 4-inch centers.  He also
proposed a set of Fast-40 configuration rules that suggested 25 meters
if the loads are spaced at least 1 meter apart and 35 meters point to
point.  The importance of using asymmetrical drivers that compensate
for the biasing effects of the terminators was clearly shown.  For
this reason an agenda item was added to see if the degree of driver
asymmetry could be tightened.  See also the following item.

A discussion of ways that might be used to decrease the intersymbol
interference suggested that there may be a significant difference
between the behavior of current mode drivers vs voltage mode drivers -
Paul Aloisi of Unitrode noted that a lower impedance terminator could
reduce the height of the long pulses.  This would give a better match
to the loaded cable impedance ~80ohms vs 110ohms for the present
terminator.  Wally Bridgewater of Adaptec floated the idea of having
the driver reduce the amplitude of long pulses after the initial bit.
This would reduce the starting point for the isolated bit transition
and result in a larger detectable signal.  There were a couple of
points that were not explored: (1) how does the driver know how large
the long pulse is due to manufacturing tolerances and (2) what sort of
algorithm should be used?

Bill's summary: The real lever is the transmission line properties and
appropriate driver asymmetry.  Rules: 8-inch spacing for cables
@20pf/ft; 4-inch spacing for PCB @40pf/ft; devices with 20pf/20pf/10pf
loading and drivers that compensate for the terminators biasing.  One
cannot compensate for poor transmission line properties by shortening
the length of the bus segment because the reflections are actually
worse in short lengths due to the lack of edge roll off.  These
capacitance value for the cable capacitance are very approximate --
Bill will do more careful measurement of the details of this property.
The rules above are based on the cables used in the testing.

5.   LVD SCSI Driver Specification Presentation (96-145r0) [Moore]

Richard Moore presented an analysis of Kevin Gingerich's earlier
model.  Richard's main point was that the requirements need to be
adjusted to more nearly represent the situation for real drivers that
need to accommodate manufacturing tolerances and other features.
Details were given in his slides.

Bill noted that his direction was what he favored and that even
further consideration needed to be given to the receiver overdrive
requirements and the tolerances on the asymmetry.  The presentation
showed a serious commitment on the part of Adaptec to support LVD
standards development.

It was shown that part of the problem with some of the driver
tolerances were due to the tolerances on the terminators.  This caused
a revisiting of the terminator specifications and resulted in an
agreement to adjust the bias range to 100-125 mV.  Also the resistive
slope range was narrowed to 100-110 ohms.  This gave some much needed
relief to the silicon side.

The question was raised concerning whether the common mode may need to
be adjusted to 100-150 ohms.  This point was not resolved and the
standard will remain as is for common mode.

Bill noted that the asymmetry should be specified more tightly than
the gray area in Richard's revised region of operation.  Asymmetry
exists even if asymmetrical drivers are not used.  Bill noted that the
asymmetry observed in the lab is caused mostly by intersymbol
interference.

Richard promised to revise a proposal for the upcoming meeting in
Florida.

6.   Standing Waves [Bridgewater]

Wally Bridgewater stated that he plans to have test results at the
meeting in June.  To be continued.  Bill noted that he hasn't observed
anything in the lab that indicates a problem, but the possibility
exists.

7.   Glitch Issues

     7.1  Releasing Bus from Active Negation [Uber]

No activity, remove from future agendas

     7.2  Digital Filtering of  Negation Release Glitches (96-158r0)
          [Lohmeyer]

John Lohmeyer presented a note on where release glitches might occur.
He promised to have a proposal that covers all these areas for the May
meeting.

     7.3  LVD Release Glitches Proposal (96-160) [Kakirian]

Shahe Krakirian of Adaptec made a presentation on LVD release
glitches.  The presentation initially proposed a new parameter, Bus
Round Trip Delay, of 150 ns.  This was viewed as unacceptable because
it would have the effect of forcing the total SCSI domain lengths to
approximately one third of the presently allowed 400 ns.  On the other
hand, one does need to accommodate release glitches by ensuring that
they are not acted on by real devices.

A new scheme was devised that fully restored the 400 ns and protects
against these glitches.  The problem case is REQ after a data phase.
Details of this proposal are in the revised presentation.  A complete
map of the requirements needed to accommodate these glitches was
created.  The consensus was to incorporate the revised map from the
proposal into SPI-2.  The timing diagram that clarifies this will not
go into SPI-2 but may be useful for the ENDL SCSI Bench Reference or
other similar documents.

We need to add something for selection timeout for data and parity.
This is John Lohmeyer's action item.

     7.4  Glitch Filtering [Ham]

This topic related to the use of a data, parity, REQ, ACK filter that
would force the receiver to "stop listening" for a period of time
after a signal transition had been recognized.  Bill Ham showed data
that illustrated a glitch in the REQ signal that could cause false
detection of REQ in illegal configurations produced by excessive
loading.  He recommended a 6 ns glitch filter exactly like that
required in SPI for negation transitions.  The silicon folks (Tak
Asami of Adaptec) pointed out that the 10 ns filter required in SPI
was not too difficult to implement because the upper bound on the
filter could be much larger than the 10 ns minimum.  In the Fast-40
case, where the assertion/ negation period may be as short as 6.5 ns,
the glitch filter would have to be much more precise.  In addition,
the situation becomes even worse for Fast-80.  The consensus was not
to require a glitch filter but to point out that it could offer
valuable protection in illegal configurations.  Implementors may wish
to incorporate such a filter if possible.

8.   Transmission Line Model (96-123 & -124) [Gingerich]

This item was not addressed because of the absence of Kevin from this
meeting.

9.   Timing Budget [Ham]

The timing budget was reviewed in light of the changes in the
measurement method for timing caused by the receiver overdrive
requirements.  Rev 6 of SPI-2 contains these revised requirements.
Time must be allowed for the signal to go from the top to the bottom
of the receiver switching range.  In addition, the assertion and
negation times for Fast-80 needed to be included.

As the transition time is highly dependent on the actual slope of the
differential waveform and this slope is determined mainly by the
properties of the transmission environment this additional time was
added to the pulse distortion skew to give a total of 3 ns.  This
effectively pushed a tighter requirement onto the drivers.

At the urging of Frank Gasparik of Symbios Logic the receiver
sensitivity for use in the timing measurements was increased to 60 mV.
This was needed because the present 30 mV level is too close to the
internal receiver offsets and will cause too much internal skew in the
receivers.  On the other hand we need to keep the 30 mV level for the
d. c. performance receiver requirements to preserve the cross talk
margin for quiescent signals.  In light of the receiver overdrive
requirements this change is almost not noticeable because the
0.25*Vtan dominates the timing measurements.

There was some debate around the exclusion of single-ended Fast-40.
Bill stated that the argument for hot-plugging might legitimize
revisiting this topic (see related topic below).  Bill proposed
removing the SE/DIFF designation from Fig 14 in the last two rows.
The assumption is that all devices will have on-chip transceivers.

It was attempted to create a budget that could be used for separate
transceivers.  The first attempt failed.  However, Richard Moore lead
a process that actually produced a possibly workable situation where
the separate drivers could have 1.5 ns skew budget and the separate
receivers could have 2.0 ns skew budget.  This latter condition will
be incorporated into the next rev of SPI-2.

The following limits were agreed on the assertion and negation
periods:  Fast-40  8 ns at the driver, 6.5 ns at the receiver; Fast-80
3 ns at the driver, 2.5 ns at the receiver.  Fast-80 will have the
same driver slew rate requirements as Fast-40.

The leakage current of existing SCSI is 10 uA, 20 uA would make the
universal driver circuit easier.  We will change this in the next rev.

10.  Hot Plugging [Ham]

Bill stated that his analysis indicates that LVD may not work properly
with hot plugging, especially the case 4 scenario.  The use of LRC and
intelligent use of REQ/ACK offset counters could mitigate the
concerns.  Another suggestion is to use a signal to quiesce the SCSI
bus segment to allow an assertion/removal.  Another is to use
repeaters or single-end Fast-40 with expanders.  A list of options was
generated for consideration at the next meeting.

The message to marketing people is that case 3 is the general-accepted
approach to take.  Case 4 can be done with the addition of isolation
interfaces (add some added cost).

11.  SPI-2 Document Review (X3T10/1142D) [Ham]

The tactical issue of what is in SPI-2 is an item for the plenary.
The working group deferred this discussion to the Florida meeting.

12.  Summary of Meeting Results


13.  Meeting Schedule

The next meeting of SPI-2 Working Group will be May 6, 1996 in Ft.
Lauderdale, FL.  Another SPI-2 Working Group is scheduled for Thursday-
Friday June 6-7, 1996, in Colorado Springs, CO at the Embassy Suites
Hotel (719-599-9100) hosted by Symbios Logic.

14.  Adjournment

The meeting was adjourned at 4:00 p.m. on Tuesday April 23, 1996.




More information about the T10 mailing list