X3T9.2/91-151

MEMORANDUM    --    09 Sep 1991

TO:	      John Lohmeyer, Chairman, X3T9.2

FROM:	      Bill Spence, TI

SUBJECT:      Possible S/E Active-Negation Driver Requirements for SCSI-3

Kurt Chan's valuable study, X3T9.2/91-130, presented the effect on termination
voltage as a function of Voh, the voltage to which the active-negation driver
raises a deasserted line, for the Boulay terminator.  My own paper, 91-126,
suggested only that the Boulay circuit of SCSI-2 would have to be modified to
survive into the era of active termination.  Pursuing the idea of not having to
modify or abandon the existing Boulay terminators, however, there appears to be
a simple approach to establishing the resulting driver requirements.

Although there is some current sinking within the Boulay circuit, it was
provided deliberately to ensure stability of the regulator.  The simple
requirement for the proper utilization of the Boulay terminator is that there
be no net inflow of current to the terminator from the bus.  Allowing for there
always being at least 2 or 3 lines low during a data transfer (BUSY and either
PARITY or a single data line per 8 bits of width), the equation to satisfy the
no-inflow requirement is simple:

2.85 - 0.5     Voh(m) - 2.85	      Voh(m) is the max allowable Voh
----------  =  -------------  where:  m is 2 for the A cable, 3 for the P cable
  110 / m	  110 / n	      n is the max number of lines pulled high

The max number of lines which may be pulled high can vary widely in different
circumstances:

P cable, all lines except the 3 or-tied lines actively negated		  -- 22
P cable, all data, parity, and clock lines actively negated		  -- 18
A cable, all lines except the 3 or-tied lines actively negated		  -- 14
A cable, all data, parity, and clock lines actively negated		  -- 10
Either cable, only the clock lines actively negated			  --  2

Active negation on only the clock lines seems to me to be a legitimate
implementation, although I have not heard of any plans for it.	In any
case, it imposes no Voh limitation at all on the active negation driver.
For the other cases, Voh(m) and the corresponding Ioh calculate out as follows
(for this purpose, positive I is taken as out of the driver):

   Cable      m        n	 Voh(m)    Ioh (ma)

     P	      3        22	 3.17	   2.91
     P	      3        18	 3.24	   3.56
     A	      2        14	 3.19	   3.05
     A	      2        10	 3.32	   4.27

Taking e.g. the last line of the table, it would produce a driver requirement
that Voh may not exceed 3.32 volts at an Ioh of 4.27 ma.  It would be valid for
a SCSI protocol chip designer to design to this limit since he would be in
control of the bus width and how many lines were actively negated.

There may be a difference of opinion as to the value of this approach.	I
believe Kurt indicated he had talked to chip designers who thought such limits
would be feasible.  Conversely, one leading chip architect I talked to feels
that a driver anemic enough to meet such limits would forfeit most of the
desired benefits of active negation.  Obviously, this subject needs working
group attention.