TO: X3T9.2 Membership                                        X3T9.2/88-150
FROM: Kurt Chan, Hewlett-Packard                             November 7, 1988
SUBJECT: SCSI Single-Ended Physical Layer Testing Proposal

I. BACKGROUND

   The single-ended physical layer of SCSI has its roots in older technology. 
As SCSI has evolved to accomodate faster, higher-density devices, some concern 
has arisen over the compatibility of old and new components and specifications.
With the higher speeds possible in SCSI-2, there is speculation that electrical
and/or configuration restrictions may be required in order to prevent bus 
misoperation. Therefore, testing is required to characterize both new and
existing systems.

II. OBJECTIVES:

  HP will attempt to determine, for both SCSI-1 and new components:

  1. Which bus configurations are most susceptible to data corruption?
     Can these be characterized into some general configuration guidelines?
  2. How much margin do we have at 6 meters (do systems fail at 7,8,9 meters)?
  3. How significant are the effects of TERMPWR decoupling on data integrity?
  4. Which of the following electrical parameters have the greatest effect on 
     data integrity:

        o Receiver voltage thresholds
        o Receiver hysteresis
        o Device input capacitance 
        o Device input leakage current
        o Adjacent signal phase, frequency, and transition times (crosstalk)
        o TERMPWR voltage
        o Terminator resistance

     where "data integrity" is defined as missing or extra REQ/ACK pulses.
  
III. SYSTEM DESCRIPTION

In order to provide controllability and observability, special SCSI "devices"
will be created which will actually be breadboards containing instrumentation.
Each device will have an analog ECL front-end connected to counter stages which
will count transitions into the logic 1 and logic 0 regions. Not only should
these counters match one another in the absence of spurious transitions, but 
they should also match the transition counts on other devices in the system.

First, standard SCSI-1 cables will be tested in the configurations described 
below (as a control). Then each new cable/connector system will be tested using
adapters to go from nonshielded low-density to the new high-density components.

The test system will consist of:

   o Up to 8 "devices"
   o A pulse generator to generate continuous REQ/ACK pulses
   o An instrument counter to count source pulses
   o A signal generator to generate "noise" on adjacent signals
   o A 1 Gsample digital oscilloscope with printer
   o A variable TERMPWR supply
   o Power supplies for the ECL, TTL, and analog reference circuitry.

IV. DEVICE DESCRIPTION

Each SCSI "device" will consist of:

   o A nonshielded device connector

   o Two high speed ECL comparators to detect transitions into logic 0 and
     logic 1 regions. The thresholds for these regions will be adjustable.
     This front end will capture glitches in the order of 5 ns, and will 
     have at least 0.2VDC hysteresis on each threshold.

   o Two 40-bit counters to log the number of transitions from the comparator
     outputs. Any extra or missing REQ/ACK pulses will show up as a mismatched
     count.

   o A discrete open-drain transistor to drive REQ/ACK. A small variable
     source resistance will be used to vary the V(ol) threshold.

   o High-speed open-collector TTL drivers to inject "noise" edges onto
     adjacent signal lines.

   o A variable lumped capacitance on signal lines to simulate parasitic 
     device capacitances.

   o A variable lumped resistance to ground on all lines to simulate receiver
     input leakage current.

   o A selectable variable-resistance termination network

   o A location for a TERMPWR decoupling capacitor. 


V. PARAMETRIC ENVIRONMENT

A "worst-case" parametric environment consists of:

   o logic 0 input threshold = 0.8V (.2VDC hysteresis)
   o logic 1 input threshold = 2.0V (.2VDC hysteresis)
   o V(ol) = 0.5V
   o TERMPWR = 4.25V (SCSI-2) and 4.0V (SCSI-1) at source
   o signals with 5ns fall time on adjacent wires
   o termination resistor networks at their maximum impedance (+10%)
   o total device capacitance = 20-30 pF (arbitrarily set initialy)
   o device leakage current = 400 uA
   o A REQ/ACK signal source at:
     (a) 1.5 MHz to simulate slower SCSI-1 (asynchronous) operation
     (b) 5.0 MHz to simulate SCSI-1 synchronous operation
     (c) 10.0 MHz to simulate fast SCSI-2 synchronous operation 

     Duty cycles will vary to force minimum REQ/ACK assertion and 
     negation times (90ns for SCSI-1, 30ns for SCSI-2).

VI. TOPOLOGICAL ENVIRONMENT

   Configuration 1:

                       +---+  Nonshielded   +---+
                       | 1 |----------------| 2 | 
                       +---+   'n' meters   +---+

   Configuration 2:
                                                  +---+         +---+
                                                  | 4 |         | 6 |
                                                  +-+-+         +-+-+
  +---+ Nonshielded          4.0 m      .25m  .25m | .25m   .25m | .25m  +---+
  | 1 |------+------[]=================[]-----+-----+------+------+------| 7 |
  +---+ .25m | .25m        Shielded           |            |             +---+
           +-+-+                            +-+-+        +-+-+
           | 2 |                            | 3 |        | 5 |
           +---+                            +---+        +---+


   Configuration 3:

                                    +---+
                                    | 3 |   
                                    +-+-+    
  +---+ Nonshielded                   |                 .25m   .25m  +---+
  | 1 |------+------[]===========[]---+---[]=========[]------+-------| 5 |
  +---+ .25m | .25m      3.0 m    0.1m 0.1m   2.0 m          |       +---+
           +-+-+                                           +-+-+
           | 2 |                                           | 4 | 
           +---+                                           +---+



VII. TESTING PROCEDURE:

1. Configuration (1): The following will be performed for BOTH standard
   nonshielded cables and new high-density components with adapters.

   Set electrical parameters to worst case and test
   IF errors THEN
      Record waveforms
      Improve each parameter one at a time to nominal and retest
      Document the parameters that improve a failing system
      Repeat with cable length shortened by 1 meter
   ELSE
      Margin test: increment cable length by 1 meter until errors occur.
   END IF

2. Configurations (2) and (3): The following will be performed for BOTH 
   standard nonshielded cables and new high-density components with adapters.

   Set electrical parameters to nominal and test
   IF errors THEN
      Record waveforms
      Document the configuration
      UNTIL errors=0 or Devices=2
         Reduce the number of devices
      END UNTIL
   ELSE
      Degrade the electrical parameters individually and retest
      IF errors THEN
         Record waveforms
         Document the configuration
      END IF
   END IF


VIII. COMPONENT REQUIREMENTS 

  The following components are required in BOTH SCSI-1 and high density
  versions:
                       -  nonshielded cable
                       +  nonshielded cable connector 
      LEGEND:          =  shielded cable
                     [c]  Alt 2 cable connector (SCSI-2 figures D-2, D-4)
                     [d]  Alt 2 device connector (SCSI-2 figures D-1, D-3)

         4.0 - 9.0 meters
      +-------------------+                        Quantity: 1 each (6 total)

        .25m   .25m
      +------+------[c]                            Quantity: 2

        .25m   .25m   .25m   .25m   .25m
      [d]-----+------+------+------+------+        Quantity: 1

          .1m   .1m
      [d]-----+-----[d]                            Quantity: 1

          2.0-4.0 m
      [c]===========[c]                            Quantity: 1 each (3 total)

      Low-density nonshielded device connectors    Quantity: 8

      Low-density to high-density adapters         Quantity: 8