DiffSense Timing

Bill Ham Bill.Ham at digital.com
Mon Nov 23 06:02:30 PST 1998


* From the T10 (formerly SCSI) Reflector (t10 at symbios.com), posted by:
* Bill Ham <Bill.Ham at digital.com>
*
Ah the perpetual discussion about transmission mode changes.

Dynamic transmission mode changes should be treated as a major fault if
it happens in an operating system.  There is no way the bus can change
transmission modes without significant disruptions since the terminators
have to change type too and in the process there is at least a
propagation delays' worth of disruption.  All this assuming that the bus
segment with the changed transmission mode will still work with the
length and loads.  Yes, devices and terminators will switch dynamically
in service if presented with a changed DIFFSENS level.  The real issue
for DIFFSENS timing is accommodating the low frequency noise AND the
time required for a connector to completely mate (at least 100 ms).
See EPI for details. This is required when adding devices to an existing
segment.  The device coming in must set ITS OWN transmission mode to
match that of the existing segment before it tries to communicate.  This
does not mean that systems should expect dynamic mode changes to work
for the segment as a whole.  We know the device is there when it
responds to inquiry or whatever.  It is not allowed to respond until the
conditions spelled out in SPI-2 are met.

We do not want devices to be changing their transmission modes unless it
is really true that a mode change has occurred (and then only if it is
an incoming device that needs to match the bus segment it will be
attached to). 

I think under these conditions that what we have now is OK.

Cheers, 
Bill Ham
Storage Bus Technical Office
Pager: 888 956-8925
Fax: 978 470-0321

	----------
	From: 	Bill Galloway[SMTP:BillG at breatech.com]
	Reply To: 	BillG at breatech.com
	Sent: 	Thursday, November 19, 1998 10:34 PM
	To: 	T10 Reflector
	Subject: 	DiffSense Timing

	I have been reviewing SPI-3 looking at the timing for driver
mode changes.
	The only timing that I can find is in section 7.2.5.2 LVD
DIFFSENS receiver.
	This section states that a device shall not change its mode
until a new mode
	has been present for at least 100ms.

	I seem to remember that the large value of 100ms was chosen to
rule out the
	possibility of 60Hz noise disrupting the system by causing mode
changes.

	Now for the problems.  There is no maximum time specified.  With
no maximum
	time,
	the initiator does not know how long it must delay after a mode
change. The
	bus is
	dead until all devices switch to the new mode.  I think the
maximum should
	be about
	300ms.

	There are no timing specifications on the terminator whatsoever.
This seem
	to
	defeat the desire that 60Hz noise not disrupt the system. I
think the
	terminators
	should obey the same timing rules as devices.

	I have talked to a few people who said it was never intended
that mode
	transitions
	happen in a real system, it is only a stocking issue. This is
NOT what
	companies are
	shipping.  Companies are shipping systems that expect hot plug
to work, even
	if it
	involves a mode change.  For mode changes to work in live
systems these
	times have
	to be specified.


	I will write up a proposal along these lines unless someone
points out the
	error in
	my ways :-)


	Bill Galloway
	BREA Technologies, Inc.
	(281) 988-0358
	BillG at breatech.com

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





More information about the T10 mailing list