[sdiy] Speed of electrons? (was mass extinction of the dinosaurs)

[Magnus Danielson]

From: KA4HJH <ka4hjh at gte.net>
Subject: [sdiy] Speed of electrons? (was mass extinction of the dinosaurs)
Date: Mon, 23 Jul 2001 17:52:39 -0400

> >The Digital Heads think in only terms of 0 and 1. The transition from 0 to 1
> >IS analog. That is where significant engineering skill is required.
> >
> >I point out in class that a nanosecond is about a light-foot.
> 
> That's for massless particles in a vacuum (~.98'/s). Electrons in copper
> wire are much slower. Years ago in Byte there was an article about DRAM
> where they through out a figure like .25" per nanosecond at 5V in copper
> traces (they used the reciprocal of 4 seconds per inch--I think). But I've
> wondered about this ever since. Does anyone have accurate figures on this?

Now, this seems like a little mess here...

We should not confuse the speed of a single electron with the speed of
change in current and potential difference. These speeds are greatly
different and gets affected by different aspects.

The speed of an electron in a conducting material will depend on the
material, the current and the physical aspects such as length and
cross-section area. If we want to look at how long time a particular
electron takes to move a meter, then we must compare the amount of
free electrons in a cross-section with the current and work it from
there. Whatever speed there is, it isn't extreme. This is the
macro-scopic speed of the electron.

The micro-scopic case is the movement of an electron from one atom to
another, and the triggering of the movement of another electron. This
microscopic movement is much, much quickier.

The middle case is the percieved transient movement of changed current
and potential difference. This movement depends largely on the
capacitance and inductance which the signal "feels" where it is. The
speed can vary over several octaves. It is not uncommon to have a
speed of 2/3ds of the speed of ligth as a signal goes down a fiber or
coax. This speed is also seen on the surface of a PCB. The signal
speed INSIDE a PCB is lower. There are also means of slowing down a
signal further, by adding more capacitance for instance.

On Semiconductors, there are several factors contributing, the
resistance and conductance for the long wires are among the few
parameters to play with. The receiver capacitance plays a role. Adding
inverters/buffers along the ride in a distributed maner will
effectively help to cut the delays. Use of copper over traditional
aluminium cuts resistance but required an investment in research on
manufactoring since copper will migrate into silicon unless stopped,
so that stopp had to be developed. The wires is usually wrapped in
siliconoxide, but other dielectrics (including air) is being tested in
order to cut time there. Recently have geometrical exercises given us
diagonal traces which works on cutting length in itself.

So, saying which speed it is, is not as trivial as you would like
it. Speed estimates in deep-submicron chips is a research topic on its
own rights.

Cheers,
Magnus - delivering an answer with just a thad of delay