Adaptec termination problem...

[Mike Bilow]

Grover Davidson wrote in a message to Mike Bilow:

 GD> I could be WAY out on this, but if memory serves me right,
 GD> the device that is FARTHEST away from the controller should
 GD> be terminated, regardless of its address. Termination,
 GD> preferably ACTIVE termination, belongs at the far end away
 GD> from the controller. Otherwise, the electrical
 GD> charictaristics that termination is designed to achieve
 GD> don't occur.

Generally, yes.  There are complications when you switch the bus midway from
wide to narrow, which happens on the 2940xW, but the principle is right.

 GD> I have a 2940W with:
 GD>    0:IBM Starfire 4.5GB Fast/Wide 
 GD>    1:Conner 540MB Fast
 GD>    2:Syquest 270MB
 GD>    4:IBM 8x CdRom (from an RS-6000)
 GD>    6:Exabyte 8200 8mm tape drive

 GD> My Wide cable has 1 activer terminator on it at the far end,
 GD> and my IBM drive is also plugged in at the far end.

 GD> My narrow bus has things plugged in in the order listed with
 GD> the higher addresses being farther from the contoller. My
 GD> termination on the narrow bus is supplied by the Exabyte
 GD> with passive termination.

In your configuration, the controller is the physical end of the high half of
th wide bus, but is in the physical middle of the low half of the wide bus
which is also the whole narrow bus.  As a result, the correct configuration is
for the controller to terminate the high half but not the low half, and then
the wide and narrow devices physically farthest from controller are terminated.

 GD> I have never had a problem with that layout. I have know a
 GD> lot of people who have had problems with their scsi system,
 GD> and about 90% of those were due to impropper termination.

It sounds as if you did it right.  If the bus is underterminated, this will
cause ringing that will result in data errors, parity errors, and possibly even
lockups.  If the bus is overterminated, it will usually work, but will support
less than the rated number of peripheral devices.
 
-- Mike



To Unsubscribe: send mail to majordomo at FreeBSD.org
with "unsubscribe aic7xxx" in the body of the message