Minutes SCSI signal modeling study group (SSM)

Dwallace50 at aol.com Dwallace50 at aol.com
Sun Jun 20 18:19:24 PDT 1999


* From the T10 Reflector (t10 at symbios.com), posted by:
* Dwallace50 at aol.com
*
This was the initial meeting to address the general subject of modeling for 
parallel
scsi. The attendees were;

Name                   Company             Email Address
Paul Aloisi            Unitrode                aloisi at unitrode.com
Larry Barnes         LSI Logic               larry.barnes at lsil.com
Jie Fan                 Madison Cable       jfan at madisoncable.com
Jonathan Fasig     Western Digital       jonathan.l.fasig at wdc.com
Bill Ham               Compaq                bill_ham at ix.ntecom.com
Tariq Abou-Jeyab  Adaptec                tajeyab at corp.adaptec.com
Martin Ogbuokiri   Molex                    mogbuokiri at molex.com
Richard Uber         Quantum              richard.uber at quantum.com
Dean Wallace       QLogic                 d_wallace at qlc.com

1.  Introduction
2. Attendance
3. Agenda development
4. Overview of effort
5. Presentations
    5.1  Termination model (Paul Aloisi Unitrode)
    5.2  Cable model (Tariq Abou-Jeyab Adaptec)
6. Output of group
7. SFF backplane
8. Components to be modeled
  8.1 Cable assemblies
    8.1.1 Cable media (bulk cables)
    8.1.2 Connectors (on cable)
    8.1.3 Transition region
  8.2 PCB's
    8.2.1 Host
    8.2.2 Target
    8.2.3 Interconnect (e.g. backplanes)
  8.3 Connectors
  8.4 Termination
  8.5 Transceivers
    8.5.1 Chip packages
    8.5.2 Access to actual measurement points.
9. System configurations
10. Data patterns
11. Data rate
12. Verification/correlation
  12.1 Physical measurement points
13. Definitions
14. Tools
  14.1 Behavioral
  14.2 Circuit

Presentations

Termination model: Paul Aloisi described a model for an integrated terminator.
This particular design has an output impedance that is flat out to 400MHz. 
Paul
also pointed out that there is considerable difference between output 
impedance
versus frequency for other terminator makers. Paul also reviewed a terminator 
model for a regulator and discrete resistor stacks (power termination).

Cable Model: Tariq Abou-Jeyab Adaptec described a Pspice model that he was
developing. Tariq tried to build a lumped model using discrete components to 
model
the lumps and discrete components to model the skin effect. The simulation 
results
did not agree with measurements. The group decided that it is very difficult 
to model
a cable using this method and tools that allow more flexibility in entering 
measured
parameters and calculating loss due to skin effect are probably better suited 
for
lossy cable simulation

The modeling of the skin effect is being examined by Larry Barnes through some
elaborate schemes. The skin effect is currently one of the most difficult 
things to
model accurately. It is difficult also to know how much of the measured loss 
comes
|from skin effect and how much comes from dielectric losses.

6. Output of the group.

The study group will produce two types of output, 1) reports to the scsi 
working

group, 2) and a committee document containg the technical details. Also a
site
that will contain simulation models for complete system simulations.

7. SFF Backplane

Bill Ham noted that work was starting on a SFF backplane.

8. Component models.

8.1.1 Cable assemblies
These consist of the media, connectors, and a transition region from the 
connector
terminaiton to the media.

8.1.2 Cable media, much discussion centered around how to most accurately 
model
the cable. Lossy transmission line models must be used. The following mesured 
parameters are needed to construct an accurate cable model. L(f), C(f), G(f), 
R(f) and
potentially the scattering (s) parameters. The frequency range is from the 
fundamental of the datarate to 0.59/trise (20 to 80 percent number presented 
by Larry). The cable parameters are represented by matrices
Rm,n(f) is the dc resistance and the skin effect resistance
Gm,n(f) is the dielectric loss.
Lm,n(f) is the mutual inductance and the self inductance
Cm,n(f) is the capacitance both coupling and to ground.

The size of the matrix depends on how many conductors are used in the coupling
calculations. The two conductor model was worked out, but there was some 
disagreement as to how many conductors have to be considered for crosstalk. 
Larry Barnes took the action item to see what this number is, his first 
thought was the five closest pairs. When Larry does this calculation he will 
consider conductors that have a mutual inductance that is within 10% of the 
self inductance, anything less than this
will be ignored.

Jie Fan agreed to present measured numbers for the next meeting that can be 
used in the simulation model. These numbers will include the frequency 
effects from 40MHz to 400MHz at 40MHz steps. Jie will provide the measurement 
methods and the measurements for AWG 30 cable.

Skew will be accounted for in the model by adjusting the cable length in the 
simulation to give the proper skew numbers between pairs. Only pair to pair 
skew will be considered, conductor to conductor skew will not. The other 
method for adding in skew would be to adjust the parametric numbers in the 
model, but this was decided against.

Verification of the simulation model with measured results should be done by 
the following:
1) Input step response with TDR measurements, this will give the impedance of 
the model.
2)Attenuation versus frequency.
3) Crosstalk.

There was some discussion on how to do a simulated TDR and Martin Ogbuokiri 
will give a presentation on this next time.

8.1.2 Connectors on cable assemblies

Were not discussed due to lack of time.

8.1.3 Transistion region

Not discussed due to lack of time.

8.2 PCB's

8.2.1 Host

There was some discussion as to how many connectors should be considered on 
the host board, the number agreed to is two wide connectors. Tariq agreed to 
do a host board model.

8.2.2 Target

Dean agreed to do a targe board model.

8.2.3 Interconnect board

Not discussed due to lack of time.

8.3 Connectors

Do good connector models exist? Only models available are for mated 
connectors. Molex has connector models for the VHDCI and SCA-2.


8.4 Termination

See Paul Aloisi's presentation.

8.5 Transceivers

Question of how to model due to the proprietary nature of the designs.

The rest of the agenda was not covered due to lack of time. The next meeting
will be July 29. A more complete document that will be updated after each 
meeting will be uploaded to the website, the document nubmer is 99-204r0.

Dean Wallace
QLogic Corporation



  



Larry

*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at symbios.com





More information about the T10 mailing list