OT: Re: How does gigabit ethernet work?

From: Eric Smith <eric_at_brouhaha.com>
Date: Fri Feb 12 18:18:59 1999

On 11 Feb 1999, Eric Smith wrote:
>*always* an analog signal. Cables won't pass digital signals. If you put in

Max writes:
> But the pind of a microchip are also cables, of sorts. What's the
> difference?

1. They aren't multiple meters long.

2. They actually do have transmission line effects. If you're not careful,
    there can be enough ringing on the signals (due to lack of proper
    termination) to cause undesired behavior.

If you look at bus signals on a 33 MHz bus with a 100 MHz scope, they look
fairly square (unless things are badly designed and/or broken). Now look
at those same signals with a 1 GHz scope. The difference is *amazing*.

At best, all you can do with high speed stuff is get a good enough
*approximation* of the desired digital signal. And the higher the speed
the harder it is to make that approximation good enough.

When you're talking about "digital" signals above 500 MHz (as would be
found on gigabit ethernet twisted pairs), it is very hard to convincingly
describe them as digital. Sure, they are enocded from digital bits, and
the are decoded back into digital bits at the other end, but what happens
inbetween is analog (and basically black magic at that).

Eric
Received on Fri Feb 12 1999 - 18:18:59 GMT