Minutes of Ultra-SCSI Ad-Hoc Meeting

R. S. Shergill rss at berlioz.nsc.com
Fri Apr 8 14:59:49 PDT 1994

                                                                Doc No. __________

Minutes of the X3T10 Ultra SCSI Ad-Hoc Meeting

Held March 31, 1994, at Milpitas, California.

Chaired by:          Jim McGrath, Quantum Corp (AM); and
                     John Lohmeyer, AT&T/NCR (PM).
Hosted by:           Jim McGrath, Quantum Corp.
Minutes recorded by: Robbie Shergill, National Semiconductor Corp.

John Lohmeyer brought the meeting to order at 9:30 AM and asked each person in 
attendance to introduce themselves. An attendance sheet was also circulated.


Name               Company           Phone         e-mail
----               -------           ------        ------

Norm Harris        Adaptec                         nharris at adaptec.com
Larry Lamers       Adaptec           408-945-8600  ljlamers at aol.com
Mark Knecht        AMD               408-235-8043  mark.knecht at amd.com
Joseph Chen        Cirrus Logic      510-226-2101  chen at cirrus.com
Dave Weber         Cirrus Logic      510-226-2367  
Dale Smith         Cirrus Logic      510-226-2151
Bill Gintz         Conner            408-456-3648
Louis Grantham     Dallas Semi.      214-450-8110  grantham at dalsemi.com
Michael Smith      Dallas Semi.      214-450-0457  msmith at dalsemi.com
Bill Ham           DEC               508-841-2629  ham at shwsys.enet.dec.com
Robert Liu         Fujitsu           408-894-3790
George Rasko       IBM               408-256-2427
Aaron Olbrich      IBM               408-256-4049
Ron Roberts        Maxtor            408-432-4322  ron_roberts at maxtor.com
John Goldie        National Semi.    408-721-2075  cjfgsc at tevm2.nsc.com
Todd Nelson        National Semi.    408-721-7846  ctwnsc at tevm2.nsc.com
Robbie Shergill    National Semi.    408-721-7959  rss at berlioz.nsc.com
John Lohmeyer      NCR/AT&T          719-573-3362  john.lohmeyer at ftcollinsco.ncr.com
Ting Chan          Q-Logic           714-668-0487  t_chan at qlc.com
Jim McGrath        Quantum           408-894-4504  jmcgrath at qntm.com
Chris Millsaps     Quantum           408-894-4269
Farbod Falakfarsa  Quantum           408-894-4066
Brian Davis        Seagate           408-439-2137  brian_davis at notes.seagate.com
Mike Fitzpatrick   Seagate           405-324-3478
Dean Wallace       Silicon General   408-898-8121
Steve Finch        Silicon Systems   714-573-6808  5723283 at mcimail.com
Vit Novak          Sun                             vit.novak at sun.com
Pete Tobias        Tandem            408-285-9913  tobias_pete at tandem.com
Kevin Gingerich    TI                214-997-3378  4307725 at mcimail.com
Paul Aloisi        Unitrode          603-429-8687  aloisi at uicc.com
Duncan Penman      Zadian            408-293-0800

John briefly described the ANSI rules governing an ad-hoc meeting. 

Jim McGrath next presented the proposed agenda for this meeting. The agenda was 
discussed briefly and adopted as follows:

         Introductions                   - McGrath
         Date and Place of Next Meeting  - McGrath
         Purpose of this SSWG            - McGrath
         Cable Plant                     - Ham
         Timing                          - Adaptec

         Ad-Hoc Meeting:     5/5/94, at Milpitas; host: Quantum.
         SCSI Working Group: 5/18/94, as part of Plenary Week in Harrisburg, PA.
         Ad-Hoc Meeting:     6/2/94, at Milpitas; host: Quantum.

Larry Lamers was not sure about the need for the May 5th meeting. Jim McGrath 
asserted, however, that a meeting before the next plenary week may be necessary 
because if anyone has any objections then they have to be brought up as soon as 
possible Since design work is already taking place, McGrath warned that significant 
changes after the May 18th meeting will be very unpopular and better have very good 
and well supported basis for them.

It was pointed out that the term "Ultra SCSI" is a trademark of Ultrastor, so we'll 
have to come up with a different name. Turbo SCSI has been suggested. This issue 
will be further discussed in the plenary. In this meeting John Lohmeyer suggested 
FAST-20 and the group was basically in agreement - other than Bill Ham's comment 
that FAST-40 may be better (i.e., assume wide applications).

Bill Ham presented the results of further work on his test setup.
The test setup has been improved since the last time. It now has real single ended 
loads; shorter stub lengths; software for disabling active negation on data lines; 
and 40mt/s diff xcvrs.

On this setup Bill Ham has demonstrated successful operation at 40mt/s over 20m.
However, Bill emphasis in the presentation was on the "ultra-SCSI 1" setup results. 
Essentially, with a fully loaded system, Bill's system is showing about 39ns window 

A critical parameter for such performance is the slew-rate. Steve Finch wants the 
slew rate spec to be the same for Fast and Ultra operations because the same silicon 
is going to accommodate both modes; but John Lohmeyer and Larry Lamers want to allow 
faster slewrate if we can also accommodate "Ultra-2" operation (40MT/s) with it.
Larry wants minimum 350-400 mv/s and 800mv/s maximum; however, Bill Gintz sees 
trouble before you get to 800mv/s  due to worst case system conditions.

Bill's data shows very little jitter, but Paul Aloisi pointed out that this is with 
the TI silicon that does not have compliant threshold levels as per the new spec. 
Bill agreed that the jitter will get a lot worse with threshold movement, but still 
not a problem in his opinion.

Bill agrees that he needs to do take more data with the receiver moved further down 
the cable, but bottomline is that adding real SCSI loads doesn't degrade things 

Bill Ham presented a set of recomendations based on his data. Each item was 
individually debated.

There was agreement on 1.5 meter cable length. However, Paul Aloisi pointed out that 
the SPI spec allows 32 devices on the bus. The group was in agreement that we don't 
want to attempt more than 16 devices.

Bill had seen the timing window shift with longer stub lengths and as a result he 
was suggesting moving the position of clock to favor setup time by 5ns. This was 
deferred because the group needs to better understand this phenomenon. Finch 
suspects that it may be caused by the TI chips' threshold levels.

Bill suggested minimum rise and fall time of 4ns and maximum 8ns (on a specified 
test load). John Lohmeyer wanted the maximum time spec removed because the 2:1 ratio 
is too restrictive for silicon. This was agreed to. 

Dale Smith suggested using 600mv/s as the maximum slew-rate. Bill Gintz agreed as 
long as this is on the test circuit which is very lightly loaded as compared to the 
worst case system configuration. All agreed to this number.

Jim McGrath wanted to specify the worst-case capacitive loading. Bill Ham guessed 
that worst-case capacitance is around 500pf. The group agreed that we should add a 
worst case test load; the question is where should it go? McGrath wanted it added to 
the spec itself, not just an annex, because it helps the silicon guys. Lohmeyer 
wanted to survey the chip specs and make sure this worst case spec is as close to 
that as possible. Group's feeling was that it has to be less than 500pf.

Bill's data showed limited signal swing that leaves no room for the hysteresis specs 
agreed to at the last meeting. Bill wanted the hysteresis specs tightened up. This 
was not well received since the hysteresis and threshold levels were carefully 
chosen at the last meeting. Robbie Shergill asked why not tighten the environment 
instead? For example, the cable length and the number of devices could be decreased. 
Larry Lamers rather tighten up the environment also.

Kevin Gingrich wanted the system environment better defined before some of the 
outstanding silicon issues can be resolved. Therefore, it was decided to come up 
with the following lists of outstanding issues at the chip level and at the system 
level. All participants are encouraged to focus on these issues.

List of Outstanding Issues:

A)  Sysyem Level Issues:
  1. Maximum number of devices.
  2. Stub length and spacing.
  3. Maximum load capacitance and the worst case test load circuit if different from
     the current SPI circuit.
  4. Termination options allowed
  5. Allow mixed configurations of 8- and 16-bit devices?

B) Chip Level Issues:
  1. Low-level voltage levels, i.e. hysteresis levels.
  2. Maximum slew rate of 600 mv/s ?
  3. Test circuits
  4. Maximum Voh
  5. Maximum Cl the chip drivers can handle
Bill Gintz next showed his data based on a Fast SCSI setup (3m). Bill showed that 
actual logic trip levels vary with supply also, as well as the amount of hysteresis 
varying with process and temp. Plus, he felt that some amount of glitch filtering 
may need to be done. Bill showed the difference in signals depending upon the 
termination used. He also mentioned the dependence on data pattern but he cant 
define this very well. Bill Gintz's message is to limit the system envelope - "can't 
be a smorgasbord at fast rates".

At the end, Bill Ham showed differential system data. Running at 40mt/s Bill's 
system had 14ns of time window available.

There was no discussion concerning the timing numbers that have been agreed to at 
the last two meetings.

The meeting was adjourned at 3:30 PM.

More information about the T10 mailing list