X3T9.2/88-087                                              Page 1


          Digital Equipment Corporation                       July 18, 1988

          +-+-+-+-+-+-+-+
          |d|i|g|i|t|a|l|
          +-+-+-+-+-+-+-+

          To:       X3T9.2 Membership

          From:     John A. Morse

          Subject:  TERMPWR Requirements for SCSI-2



          I propose that SCSI-2 require all devices to  supply  termination
          power,  not  just  initiators.   Specifically, I propose that the
          first paragraph  of  section  4.4.3  in  Rev.   4  of  the  draft
          specification be replaced by the following:

               All SCSI  devices  shall  supply  terminator  power  to  the
               TERMPWR  and  TERMPWRB pin(s).  This power shall be supplied
               through a  diode  or  similar  semiconductor  that  prevents
               backflow of power to the SCSI device.

          This proposal has come about  after  extensive  discussion  among
          hardware  and software engineers at DEC.  Anticipated application
          of SCSI at DEC is for small systems.  Storage expansion for  such
          systems  will  often  be  provided  in  the form of free-standing
          enclosures having their own power supplies, typically  containing
          one  or  two  SCSI  devices.   We  believe our needs are far from
          unique, and in fact reflect a typical pattern of  usage  of  SCSI
          devices that will be found throughout the computer industry.

          The proliferation of stand-alone SCSI  expander  boxes  makes  it
          desirable  that  the  SCSI  devices  themselves supply terminator
          power to the bus for the following reasons:

               1.  Independently  powered  boxes  should  power  their  own
                   terminators.   This provides more solid termination than
                   depending on TERMPWR being  supplied  by  the  initiator
                   down  the  full  length  of  the SCSI cable, and is less
                   noise  prone.   It  also  provides  protection   against
                   glitches  on  the SCSI bus if the system is powered down
                   while the drive box still has power.

               2.  Given that independent boxes need to  supply  their  own
                   termination  power, it is far easier to do this right at
                   the drive electronics, than to try to attach to the SCSI
                   bus  at  some  other  point.   Typically,  at the device
                   electronics, it means a little bit of etch,  one  diode,
                   and  a  currently  limiting  device (fuse or themistor).
                   That is under $1.00 in cost.
          X3T9.2/88-087                                              Page 2


                   In a typical expander box, the  SCSI  connnection  is  a
                   pair  of  Champ connectors on the bulkhead, and a simple
                   length of ribbon cable starting at one  Champ,  visiting
                   each  drive  in  turn  with the appropriate dual in-line
                   connector, and ending at the  other  Champ.   To  inject
                   TERMPWR  into  this  cabling  scheme may require an etch
                   board for a distribution panel, brackets, added assembly
                   time,  etc.  This can easily add $20.00 to $50.00 to the
                   cost of each expander box.

               3.  We estimate that somewhere between 20% and  50%  of  the
                   SCSI  devices that DEC will use in the next 3 years will
                   be  installed  in  these  independently  powered  boxes.
                   Since  the  device can supply TERMPWR at less than 1/5th
                   the cost of external wiring, it is economical to require
                   EVERY  device  to  supply TERMPWR even if it is actually
                   REQUIRED only in one case out of five.

               4.  Our experience  has  been  that  there  is  an  economic
                   advantage  in reducing the number of variants of a given
                   component that we must qualify, document,  and  maintain
                   in  inventory  (manufacturing,  and service spares).  In
                   this case, that advantage far outweights any slight cost
                   savings  that  might be realized by providing 2 variants
                   of each SCSI device --- one that  does  supply  TERMPWR,
                   and one that does not.


          Not having been involved in X3T9.2 prior to April of this year, I
          realize  that  this issue may well have been thoroughly discussed
          and settled already, although a scan  of  the  document  register
          would  not indicate that.  Even so, I believe that the experience
          that we are gaining as system  integrators  indicates  that  this
          proposal has considerable merit.  I would like it to have serious
          consideration at the August Plenary meeting in Colorado Springs.