Arbitration and Data Out bus phase similarities?

John Lohmeyer lohmeyer at
Wed Aug 11 08:52:01 PDT 1999

* From the T10 Reflector (t10 at, posted by:
* John Lohmeyer <lohmeyer at>

The situation you describe might be a problem if SCSI were implemented as a
pure state machine.  However, much to the alarm of the purists, SCSI cannot
be implemented as a pure state machine.  One of the rules to begin
Arbitration phase is that the device must first detect Bus Free phase (BSY
and SEL both false for at least 400 ns).  This prevents the situation you
describe (and it means that SCSI chips are a bit more complicated than pure
state machines).

Give me a call if this explanation isn't clear -- I'd be glad to talk further.


At 8/11/99 07:45 AM , you wrote:
>Just a quick question.  I just noticed that the Arbitration and Data Out
>phase look the same (looking at BSY being asserted and SEL, C/D, I/O,
>ATN, MSG being deasserted).  If the situation occurred where we were in
>the middle of the Data Out phase (synchronous) and a target (call it
>targ1) has REQ'd up to the max offset and is having to waiting an amount
>of time for the ACK's.  In this time frame a second target (targ2)
>decides that it is going to try and reselect the initiator.  If it looks
>at the bus at this point could it confuse the Data Out phase for an
>arbitration phase and drive BSY and it's ID onto the bus.  Additionally,
>if this can occur ,then if the Initiator starts to respond with the
>ACK's and driving data onto the bus at the same time then there could be
>an unknown on the lower 8 bits of the data bus.  I'm not even sure how
>often (if at all) this situation could present itself, but what would
>the proper way to handle this be?  Thanks.
>Art Rodgers

John Lohmeyer                  Email: lohmeyer at
LSI Logic Corp.                Voice: +1-719-533-7560
4420 ArrowsWest Dr.              Fax: +1-719-533-7183
Colo Spgs, CO 80907

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

More information about the T10 mailing list