LVD SCSI Common Mode Resistance
HAYNES_DENNIS/error_bbc////////HPMEXT1/DENNIS#b#HAYNES#o#TUC#o#06 at u2.bbrown.com
HAYNES_DENNIS/error_bbc////////HPMEXT1/DENNIS#b#HAYNES#o#TUC#o#06 at u2.bbrown.com
Fri Jan 19 15:51:00 PST 1996
* From the SCSI Reflector, posted by:
* <HAYNES_DENNIS/error_bbc////////HPMEXT1/DENNIS#b#HAYNES#o#TUC#o#06 at u2.bbrown.com>
I have been working on equations to describe
the relationship between the various parameters
on the three resistor terminator for Low Voltage
Differential SCSI. I will show the results at the
next working group session Jan 29/30.
One of the relationships I have found is that
Rcm = Rdiff/4 x Vref/Vb
where
Rcm is the desired common mode impedance
created by the terminator.
Rdiff is the desired differential
impedance created by the terminator
Vref is the power supply at the top of
the three resistors that make up the
terminator
Vb is the desired bias voltage generated
by the terminator
If this equation is met the terminator
can be implemented with a single supply.
If the equation is not met, then two supplies
will be needed.
When I use the worst case numbers for the
various elements in the equation I get a higher
value for Rcm than we have discussed before.
Rcm = 115/4 x 2.5/0.1 = 719 ohms
This says that unless we can raise the Rcm value
to 719 ohms we will need to go to a more complex
implementation. The circuit can be realized with
two voltage sources, on at the top and one at the
bottom of the resistor string.
We now have models that give us quantifiable
effects on the values of Rdiff, Vb and Vref. So
far we have not gotten similar models that would
give us insights on the impact on system performance
as the value of Rcm is changed.
If any of you have resources that you can
determine the implications of changes in
Rcm on system performance please use them.
If we unnecessarily keep the value of Rcm too low
it will raise the cost of termination without
providing a system benefit.
Dennis Haynes
Burr-Brown Corp.
Tucson, AZ
Phone: 520 746 7262
EMail: haynes_dennis at bbrown.com
More information about the T10
mailing list