Typical TERMPWR Loads...

Trimm Industries trimm at netcom.com
Tue Nov 7 18:45:21 PST 1995


ljr at austin.ibm.com asks:

The SCSI-2 standard requires initiators to be capable of supplying terminator 
power.  For the A cable, the requirement is 900mA minimum current capability; 
for the B cable option an additional 1500mA is required.

The SCSI-3 SPI (rev 15a is the one I'm looking at) also requires initiators to 
supply termpower.  In this case, the specification is for 1500mA minimum with 
no distinction made for A or P cable options.

For adapters that implement multiple SCSI ports, this seems like a significant 
problem.  Consider a three port adapter.  According to these specifications, 
this adapter must be capable of sourcing 4500mA minimum just to the TERMPOWER 
lines of it's three buses.  For a PCI rev2.1 compliant adapter, the maximum 
total power allowed for the entire adapter is 25 watts.  When the three busses 
are drawing current at the SCSI-3 spec'd minimum capability, there is only 
500mA available to drive all the adapter logic, assuming your system can source
25 watts to an adapter continuously.  There are also trace sizing 
considerations on the adapter itself.

Now I realize the typical case is below the 1500mA minimum capability. 
However, if I'm charged with verifying a claim of compliance to SCSI-2 or 
SCSI-3, I would be forced to load termpower to these levels, right?

What ARE typical current levels for TERMPOWER in an 8 and a 16 bit environment,
anyway?

---- snip snip ----

We at Trimm reply:

We did some testing with a 7 channel RAID controller, with 7 wide
single ended ports.  The controller had to power 14 terminators
all by itself, and it has a separate power connector for TERMPWR.
Writing all ones to all 7 channels simultaneously (confirmed
by a scope and analyzer), the current probe fluctuated between
3 and 4 amps, with occasional transients twice that high.  This
had implications for our power distribution system, of course.
The biggest problem you face is that you can't allow those insane
current transients to affect either your system VCC or the 
TERMPWR lines.  I can imagine this would be a challenge on a 
PCI card where one can't physically mount large capacitors to
supply bulk decoupling of the output stage of your switching
regulator for TERMPWR.  Also note that current requirements can
go from 0 amps to 8 amps in a few nanoseconds, worst case. 

Gary M. Watson                              
Electronic Engineer                         Internet: trimm at netcom.com
Sigma-Trimm Technologies, A Division of Robroy Industries
350 Pilot Road   
Las Vegas, NV 89119                         Phone: (800) 423-2024 x2115 
** Enclosures for SCSI, RAID, Fibre Channel, and SSA **




More information about the T10 mailing list