SCAM and Differential

"Permanent address gardner" at "Permanent address gardner" at
Sat Oct 1 13:41:34 PDT 1994

I discussed SCAM and the properties of differential signals with Bill
Ham over a year ago.  Based on both engineering design knowledge and 
what Bill has seen/measured in the lab, SCAM should work just fine in
differential.  When a signal is released, it may take a while (relatively
speaking, perhaps a few microseconds) for the terminators to pull it to
a negated state, but it will occur.  And since the SCAM protocol is totally
asynchronous it will work, albeit perhaps slightly slower than in a singel
ended environment.  Oh yes, the receivers need a little hysterisis to
avoid oscillation on the slow negation edge, but apparently existing
receivers already have this.

So, differential SCAM should work, whether it is practical is another matter.
Everyone will make up their own mind as to whether SCAM is useful in a
differential environment.  However, to the best of my knowledge every single
SCSI protocol chip does not provide sufficient control of differential
transceivers to implement SCAM.  In particular they are missing the ability
to disable active negation and operate in an acctive-assert / passive-release
/ wired-or mode.  Thus SCAM should work in differential, but there is no
existing protocol silicon that can implement it, and I'm not aware of anyone
developing any.

Ed Gardner

More information about the T10 mailing list