Re: Triton DMA

Joseph H. Buehler (jhpb@sarto.gaithersburg.md.us)
30 Nov 1997 06:47:44 -0500


R.E.Wolff@bitwizard.nl (Rogier Wolff) writes:

> I'm not sure that you understand what termination is.

Nice explanation. Here's another:

When you change the voltage level on one end of a wire in a SCSI
cable, due to speed of light considerations, there is no way that the
device driving it can immediately know what load is on the other end,
and therefore what current should flow. So the current that starts
flowing in the wire has to do with characteristics of the *wire*, and
not the load on the other end.

This can only last so long. Eventually the signal change hits the
other end of the wire. Termination involves putting a load on the end
of the wire that matches the characteristics of the wire, so that the
current that started flowing when the driver changed the signal is
indeed what was needed all along.

If there is a termination mismatch, a reflection will occur. The
current in the wire *has* to match what the load wants, when the
signal reaches the load. But it doesn't, so you get a signal
propagating back down the wire from the load towards the driver. It
can take several back-and-forths to reach something resembling steady
state.

The devices on the SCSI bus see all the back-and-forth. The problem
is that this can cause them to sense false bit changes, or bit changes
at the wrong times. The net effect is as though there is noise on the
line: which causes all the various problems that you get with improper
SCSI termination.

This effect has various uses. For example, you can use it to figure
out where there is a break in your transatlantic telephone line: you
look at the timing of the reflections and can figure out how far down
the line the break is.

In high power electronics, like radio transmitters, the reflections
can have enough power that they will destroy the output stage of the
transmitter.

Joe Buehler