RJH wrote:

On 27/01/2018 09:25, Roger Hayter wrote:
RJH wrote:

On 26/01/2018 21:15, Roger Hayter wrote:
snip
Don't forget that high speed digital signals or
*not* binary at the physical layer, but more like RF digitial signals,
multi-level and phase dependent. !0GBe for instance.

JOOI, in what way are they not binary?

They are binary as far as the computer at each end is concerned, but
they are carried over the copper pairs (or fibre, but we are talking
about gold contacts!) as multi-level signals, effectively analogue
signals.

http://www.ieee802.org/802_tutorials/99-July/mastut.pdf


Ooooh, didn't know that! Not sure what I thought happened, mind.

Presumably it gets unpacked at either end with no change to the data -
is the 'cost' some very minor latency perhaps?

The paper is way over my head, but thanks anyways. References to FEC
seem to be relevant, but I'm not sure.

The main point I was making was that we now get the very most out of our
copper for digital signals, to the point that they are as critical for
good contacts as low level analogue signals. In fact, more so because
of transmission line effects. Any intrinsic robustness of digital
signals has been traded for frankly incredible (to someone trying to get
up to 30MHz out of a pentode fifty years ago) bandwidths.

--

Roger Hayter