h2g2 The Hitchhiker's Guide to the Galaxy: Earth Edition

I think the PCI bus may use something similar to this, but I'm not familiar enough with the physics to be sure. It's certainly different to the way that most computer buses work.

Bear in mind that I'm not an electronic engineer when reading the following, so it may be a bit, well, wrong.

Normally on a silicon chip we have massive powerful transistors (well, relatively massive and powerful, compared to all the other ones in the chip) to drive the level of wires connected to a chip either to '1' or '0' (typically represented by something around 3V and 0V respectively). These things are hefty because they're trying to raise the voltage of the wire as quickly as possible. This has the effect of sending a wave front (often called a 'spike' in the industry because of how they look when you observe them on an oscilloscope) down the wire.

If you don't do something about this spike, then as soon as it gets to the end of the wire, it will bounce off it and come right back, and rattle back and forth along the wire before settling down. This phenomenon is called 'ring', and can cause havoc, and certainly slows things down (you have to wait for everything to settle). So what we usually do is put something akin to a buffer at the end of the tracks - usually a resistor designed to soak up the energy of the wave front as it gets to the end of the wire.

Of course this is all a bit energy inefficient - pump a lot of power into the wire, and then absorb most of it at the other end.

PCI takes a different approach. I gather that it actually uses the wave front as the signal, and not the voltage level left after the wave front has passed. And I think it even lets the thing reflect, although I can't remember why. I'm told this means you can drive the system much faster without having to put in the huge amounts of power the older techniques would have required.

Don't know about PCI bus. You'll have to wait for someone more knowledgeable to answer.

What struck me, though, was the comment about the reflected wave. This may be apochryphal, but....

When Texas Instruments were designing the first logic integrated circuits using what became known as the TTL mechanism, they were experiencing terrible signal spiking and reflection. They had no idea how to resolve the problem until one of the team was discussing the problem with a power engineer who recognised the problem as exactly what happens in transmission lines.

These cross-overs and coincidences always strike me as amazing.