HPT372 bug summary [was: RE: escalation stage 2]

Terry Lambert tlambert2 at mindspring.com
Fri Jul 18 00:28:09 PDT 2003


Harald Schmalzbauer wrote:
> Please forget that. It was because for convinience reasons I had turned the
> 80-pin ATA cables upside down. So the black was at the controller and the
> blue at the drive.
> I can't imagine that this makes any technical difference (as long as no
> slave drive is connected and there's no open end)
> But it seems the single connectors are electrical coded (again I can't
> imagine how?!?)

>From the "Computer Lore For What It's Worth" department...

In cables with free wires in them, it's very common to ground
them on only one ends of the cable (usually, the end with the
best chassis ground, which in this case would be the host), so
that the free wires act to damp noise in the environment from
effecting the other lines... basically, a Farraday cage.

It's very common to see this in old RS-232C cables, as well as
in SCSI cables, especially the long ones used in differential
SCSI, if they don't have shielding.

The reason you connect only one end is to avoid ground loops,
which you can get if you use line voltage base reference: for
example, on RS-232C, it was common to abuse the standard and
send a digital ground of 12 volts on the digital ground as the
zero voltage base reference, which gave you -11 (= +1) and +11
(= +23) for 24v telephone equipment, so that you didn't have to
have a power supply capable of putting out -11v and regulate
the +24 to +11.

If the cable cared, it probably cared because it was grounded
on the host side and not on the drive, and the drive probably
didn't hook the pins up at all, to avoid ground loops when
using cables that *did* connect the pins through.

Best case, you were connected to digital instead of chassis
ground on the drive side, and worst case, you were unconnected
on the host end *and* the drive end, and so had a nice big
linear capacitor/dipole antenna.  8-).

In other words, if a cable says to connect it a specific way,
there's probably a good reason.

-- Terry


More information about the freebsd-current mailing list