mixing LVDS and SE in same command?

Skip Jones sk_jones at qlc.com
Fri Oct 6 14:46:53 PDT 1995


Gerry,

Thanks for re-sheding the light.  Darn, Tak was getting my hopes up!

Oh well, never mind.

sj
_______________________________________________________________________________
From: Gerry Houlder on Fri, Oct 6, 1995 9:59 AM
Subject: Re: mixing LVDS and SE in same command?
To: scsi

Tak wrote:

>I was wondering what's limiting us from switching back and forth between 
>SE and LVDS.  So a HBA can negotiate with each target to go LVDS during 
>the data transfer, just like you negotiate synchronous transfer.  I think 
>that's more useful, and we almost have all the hardware to do it.  
>We will need a new message or twisting of an old message (like negotiage 
>for 25nsec cycle and go LVDS).
>This is only between the initiator and the target with a nexus.  No one
>else need apply, and consequently, all arbitration and selection processes
>are performed single ended.  Now that's a backward compatibility.

LVDS signals need two lines, both of which need to change signal levels to
operate according to differential rules. Existing SE devices will have a 
signal on one line of this "signal pair" and GROUND on the other line of that 
pair. Thus if even one old device is on the SCSI bus, differential signaling 
is impossible because the "old" device will keep those grounds on the bus ALL 
the time, not just when it is sending/receiving a command on the bus. 
Thus the choice of SE or LVDS is determined exclusively by the DIFF 
SENSE line. The DIFF SENSE line may change when device(s) are 
added to the bus or removed from the bus, but changing the tranceiver 
type at any other time is impractical.

Unless, of course, your miracle dual mode driver can successfully do 
LVDS levels even when one of the outputs are grounded? This is truely a 
miracle.





More information about the T10 mailing list